Ffmpeg-devel-irc

Download

ffmpeg-devel-irc@ffmpeg.org

May 2016

1 participants
62 discussions

ffmpeg-devel.log.20160515
by burek 16 May '16

16 May '16

[01:12:07 CEST] <cone-324> ffmpeg 03Peter Ross 07master:86e493a6ffac: avcodec: add Direct Stream Transfer (DST) decoder [01:26:48 CEST] <cone-324> ffmpeg 03Paul B Mahol 07master:f57897e498fb: avcodec: bump minor and update Changelog for recent DST addition [01:26:49 CEST] <cone-324> ffmpeg 03Paul B Mahol 07master:26eee872e8ca: avcodec/escape124: change transitions table type to int8_t [01:26:50 CEST] <cone-324> ffmpeg 03Paul B Mahol 07master:9b017ac7f1c7: doc/general: mention WSD demuxer [01:26:51 CEST] <cone-324> ffmpeg 03Paul B Mahol 07master:6543a72fcefc: doc/general: mention DST decoder [01:35:25 CEST] <cone-324> ffmpeg 03Will Kelleher 07master:964f07f68e1c: hevc: Fix memory leak related to a53_caption data [04:07:36 CEST] <cone-324> ffmpeg 03James Almer 07master:81c3ce0855f1: doc/general: update supported DCA extensions [04:07:37 CEST] <cone-324> ffmpeg 03James Almer 07release/3.0:68dcb46205c7: doc/general: update supported DCA extensions [14:34:31 CEST] <cone-581> ffmpeg 03Michael Niedermayer 07master:fd4a52e4da64: avcodec/dstdec: Fix "warning: initialization from incompatible pointer type [enabled by default]" [16:36:40 CEST] <cone-581> ffmpeg 03Paul B Mahol 07master:5350e0fc97a5: avcodec/iff: rewrite out of bounds checking in writer [17:15:33 CEST] <wm4> michaelni: can we have this in libswresample? http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavresample/audio_mix.c;hb… [17:15:51 CEST] <wm4> you get distortion on non-normalized s16 downmix otherwise [17:47:02 CEST] <michaelni> wm4, what is "this" exactly ? also i need a testcase [17:56:20 CEST] <wm4> michaelni: clipping for integer mixing [17:56:36 CEST] <wm4> test case https://0x0.st/qN0.dts [18:01:24 CEST] <kierank> wm4: have you not had complaints about aac 5.1 -> 2.0 downmixing [18:02:38 CEST] <wm4> for most of the time I had normalization enabled on downmixing, which doesn't affect this [18:06:39 CEST] <cone-581> ffmpeg 03Michael Niedermayer 07master:35c23bf67df6: avcodec/dsd: Fix "warning: ctables defined but not used [-Wunused-variable]" [18:40:57 CEST] <Daemon404> nevcairiel, had a chance to poke h2645? [19:03:06 CEST] <cone-581> ffmpeg 03Michael Niedermayer 07master:7fe81bc4f8ba: swresample/rematrix: Use error diffusion to avoid error in the DC component of the matrix [19:04:50 CEST] <michaelni> wm4, fixed the issue i found with the file, if theres something remaining please say so [19:08:13 CEST] <cone-581> ffmpeg 03Paul B Mahol 07master:5afecff1effe: avformat/adxdec: set bit_rate, fixes duration calculation [21:09:31 CEST] <wm4> michaelni: thanks, but that supposedly doesn't help if normalization is not enabled [21:13:00 CEST] <durandal_1707> Then enable it? [21:15:28 CEST] <wm4> durandal_1707: then it's too silent [22:06:34 CEST] <nevcairiel> supposedly nvidias pascal chips have a 10-bit hevc encoder with HDR support (for streaming HDR games to TVs through their game streaming platform) [22:06:47 CEST] <JEEB> funky [22:07:18 CEST] <JEEB> haasn just implemented initial SMPTE ST2084 support into mpv [22:07:36 CEST] <JEEB> to get the buzzword of the year [22:07:45 CEST] <haasn> madVR compatibility [22:27:27 CEST] <Shiz> mad VR compatibility* [22:27:58 CEST] <cone-581> ffmpeg 03Michael Niedermayer 07master:2f76157eb05b: swresample/rematrix: Use clipping s16 rematrixing if overflows are possible [22:28:38 CEST] <michaelni> wm4, fixed too i hope [23:12:09 CEST] <haasn> Shiz: but does madVR support VR yet? [23:12:21 CEST] <haasn> I should make a doom9 account and ask if madVR will have SteamVR support [23:12:44 CEST] <haasn> s/<VR>/mad VR/ [23:50:06 CEST] <Compn> its not virtual reality until it plugs into your brain, man [23:51:55 CEST] <haasn> It's not virtual reality until it can simulate quantum processes :/ [23:52:04 CEST] <haasn> QVR when? [23:53:32 CEST] <iive> evert game is virtual reality [23:53:44 CEST] <iive> it just doesn't feel so much real for us. [23:54:23 CEST] <iive> but VR headset are already changing that. [00:00:00 CEST] --- Mon May 16 2016

1 0

ffmpeg.log.20160515
by burek 16 May '16

16 May '16

[01:13:14 CEST] <neuro_sys> Hello again. Despite the timings are right, there's a slight delay between streams for this command: https://gist.github.com/neuro-sys/1d857664c5a32cabae4a2b1d0a826a60 [01:13:49 CEST] <neuro_sys> I wonder if there is such a thing like one should synchronize streams first before passing them to filters? [01:14:05 CEST] <neuro_sys> Or what might be the reason of the delay between two streams that are overlayed on top. [02:20:36 CEST] <interest1ng> hey all [02:20:47 CEST] <interest1ng> firstly, thanks for making ffmpeg, it's great :) [02:21:17 CEST] <interest1ng> however, i have encoded some videos with vp9, and seeking/scrubbing (going to another time in the resulting video) is very slow [02:21:24 CEST] <interest1ng> for example, i have used this command: [02:21:31 CEST] <interest1ng> ffmpeg -i BigBuckBunny_512kb.mp4 -c:v libvpx-vp9 -crf 10 -b:v 0 -c:a libopus BigBuckBunny_512kb.webm [02:23:02 CEST] <interest1ng> the original mp4 allows me to select any time and goes to it pretty much instantly, but the webm takes quite a few seconds [02:23:05 CEST] <interest1ng> what gives? [02:30:06 CEST] <kode54> hardware decoding versus full software decoding? [02:30:18 CEST] <kode54> improper use of keyframes? [02:35:19 CEST] <interest1ng> kode54: i'm quite new to codecs and all, but how could i fix the keyframes issue? (vp9 is most likely purely software decoded anyway) [02:44:14 CEST] <neuro_sys> Is AV_PIX_FMT_RGB24 all continous RGBRGB without any paddings? [02:45:50 CEST] <neuro_sys> Do we actually need to take AV_Frame's line_size into account when reading its data and pix_fmt is RGB24? [02:46:53 CEST] <pzich> In the docs I see "AV_PIX_FMT_RGB24: [02:46:55 CEST] <pzich> packed RGB 8:8:8, 24bpp, RGBRGB..." [02:51:48 CEST] <neuro_sys> Well then can you help me identify what's wrong with this one? https://gist.github.com/neuro-sys/dd78f1672af4d1b93036360681a1a063 [03:02:35 CEST] <neuro_sys> nevermind, it's fixed [05:20:24 CEST] <jclone> hell-o [05:20:53 CEST] <jclone> looked about the ffmpeg site, where are the gpg public keys? [05:21:56 CEST] <c_14> Listed at the bottom of the MAINTAINERS file [05:23:06 CEST] <jclone> I'm looking for the keys for checking the releases sigs [05:23:22 CEST] <jclone> not really to participate in development, perhaps later! [05:24:37 CEST] <c_14> They're probably signed by Michael's key [05:25:03 CEST] <jclone> ahh [05:26:15 CEST] <jclone> Thank you! [05:29:00 CEST] <c_14> Ah, no. There's apparently a release signing key [05:29:35 CEST] <jclone> ok where is that..sorry [05:30:06 CEST] <jclone> I thought it would for the entire org [05:32:46 CEST] <c_14> The release signing key is also listed in the MAINTAINERS file [05:33:31 CEST] <c_14> As FFmpeg release signing key [05:35:16 CEST] <jclone> thanks [05:35:21 CEST] <jclone> cheers [05:52:07 CEST] <jclone> c_14, I got it from the mit public key server, all good, fingerprint checks out. [05:52:29 CEST] <jclone> https://pgp.mit.edu/pks/lookup?op=get&search=0x611EC787040B0FAB [05:53:12 CEST] <jclone> thanks again. [06:44:55 CEST] <Admin__> hey guys.. can anyone see here why it would not respect the bit rate ? $ffmpeg_path/ffmpeg -y -i "$files" -fflags sortdts -vcodec nvenc -preset llhq -2pass 1 -b:v 1800k -minrate 100k -maxrate 2500k -bufsize:v 1800k -bf 2 -g 15 -i_qfactor 1.1 -b_qfactor 1.25 -qmin 21 -qmax 23 -c:a libfdk_aac -b:a 128k -ac 2 -f mpegts /mnt/extra/$channel/"$files".conv [06:45:21 CEST] <Admin__> basically its encoding with nvenc and it shows output should be 1800k .. but when i look at the actual bit rate its like 10Mbit/s which is the source speed [06:45:24 CEST] <Admin__> any ideas ? [06:57:04 CEST] <Admin__> anyone? [06:57:26 CEST] <anachronism> What is the CSCD encoding exactly? I can't find much info on it, other than that it seems to be related to camstudio, and the bitrate leads me to believe it's lossless and uncompressed. [07:02:06 CEST] <anachronism> Nvm, I found some info. http://wiki.multimedia.cx/index.php?title=CamStudio_Screen_Codec [09:14:50 CEST] <ehsanv> hi, i'm trying to convert a bgr565be format raw data to bmp file, all ok but in some file ffmpeg give me error : "P22.raw: Invalid data found when processing input" [09:15:13 CEST] <ehsanv> actually files are not raw format, they are only pixel data! [09:17:27 CEST] <ehsanv> my command is: ffmpeg -s 640x479 -pix_fmt bgr565be -i P22.raw P22.bmp [09:18:30 CEST] <ehsanv> + if there is a way that i can only give "width" and ffmpeg guess height would be a lot better! i tried srcw but it gave me error [09:20:25 CEST] <furq> ehsanv: you may need -f rawvideo and/or -c:v rawvideo before -i [09:25:08 CEST] <ehsanv> furq: thanks man, worked! [09:34:33 CEST] <ehsanv> sorry i disconnected [09:35:17 CEST] <ehsanv> furq: and what about input width? i want to tell ffmpeg what is width and it guess the height [09:40:10 CEST] <furq> guess it based on what [10:02:50 CEST] <Fjorgynn> Life [10:47:49 CEST] <ehsanv> furq: based on input size [11:53:18 CEST] <MrSassyPants> How to enable scene cut keyframes in libvpx / vp9 ? I can't seem to find the right doc [12:13:08 CEST] <dershark_> hey , I'am trying to build ffmpeg for tvheadend on freebsd , but I won't compile . Don't know where to start troubleshooting http://pastebin.com/L1qmJUmR [12:14:51 CEST] <furq> dershark_: what's wrong with the version in ports/pkgng [12:17:22 CEST] <dershark_> it is not 3.0.2 and I don't know if that works for tvh , but I didn't tried. [12:17:39 CEST] <dershark_> try* [12:40:47 CEST] <ap0> Hello. I am using ffmpeg co trim movies without encoding. for that I tell it a start time and an end time. Is there also a way to tell fmpeg to trim by chapters instead absolute time? [13:12:41 CEST] <ChocolateArmpits> ap0: you'll have to use a script for that, get the chapter times and use as arguments for the start and end time parameters [13:23:21 CEST] <ap0> ChocolateArmpits, too bad. But thank you. [13:23:36 CEST] <JEEB> or you can of course implement the feature yourself [13:23:48 CEST] <JEEB> since the chapters should be available to ffmpeg.c [13:24:03 CEST] <JEEB> (that's the command line application) [13:26:52 CEST] <ap0> It would take me a very long time to read myself into that project [13:27:43 CEST] <JEEB> possibly [13:56:04 CEST] <Roest> hi, anyone here that could help with a problem with libav, I'm trying to export video, basically using the example code, but for some reason it encodes about 10 seconds of video and then just skips the rest [13:56:15 CEST] <Roest> http://pastebin.com/Nrpz75bX [14:00:34 CEST] <MrSassyPants> How do you enable scene cut keyframes in libvpx / vp9 in ffmpeg? I can't seem to find it in the documentation [14:56:52 CEST] <Admin__> hey guys.. can anyone see here why it would not respect the bit rate ? $ffmpeg_path/ffmpeg -y -i "$files" -fflags sortdts -vcodec nvenc -preset llhq -2pass 1 -b:v 1800k -minrate 100k -maxrate 2500k -bufsize:v 1800k -bf 2 -g 15 -i_qfactor 1.1 -b_qfactor 1.25 -qmin 21 -qmax 23 -c:a libfdk_aac -b:a 128k -ac 2 -f mpegts /mnt/extra/$channel/"$files".conv [14:56:57 CEST] <Admin__> basically its encoding with nvenc and it shows output should be 1800k .. but when i look at the actual bit rate its like 10Mbit/s which is the source speed [14:57:00 CEST] <Admin__> anyone? [15:06:07 CEST] <jkqxz> Admin__: Bounding the qp in [21,23] likely makes your bitrate setting ineffective. [15:07:04 CEST] <Admin__> really? [15:07:14 CEST] <Admin__> do you mean that i should leave it automatic ? [15:08:51 CEST] <jkqxz> Yeah. Try removing the -i_qfactor, -b_qfactor, -qmin and -qmax arguments. [15:09:23 CEST] <Admin__> ok.. i just want to get the best quality out of it.. also -ref 1 .. i tried to put that in but it says ref is not a known option.. very odd [15:15:09 CEST] <jkqxz> Not sure the reference frame count will do anything even if you do manage to set it. [15:15:38 CEST] <jkqxz> Really if you want best quality you shouldn't be using a hardware encoder. [15:18:56 CEST] <ripmind> Hi, how can i convert a 3d to 2d? I tried: ffmpeg -i Bolt\ $2008$.mkv -aspect 16:9 -vf crop=960:1080:960:0 -vcodec copy -acodec copy bolt\ 2d.mkv [15:22:30 CEST] <iive> ripmind: look in `man ffmpeg-filter` for "stereo3d" one. it have output options for mono (left eye or right eye) [15:24:13 CEST] <ripmind> Thanks iive [15:34:19 CEST] <Admin__> no hardware encoding ? but the new nvidia m4000 cards are very good quality.. and the do lossless [15:47:19 CEST] <afox_> hello [15:53:38 CEST] <ripmind> iive: stereo3d filter is not available [16:03:18 CEST] <ripmind> How do i compile ffmpeg with stereo3d filter? [16:03:19 CEST] <Admin__> jkqxz > one more thing.. not sure why i can't get low latency enabled [16:04:08 CEST] <Admin__> $ffmpeg_path/ffmpeg -y -i "$files" -fflags sortdts -vcodec nvenc -preset llhq -2pass 1 -b:v 1800k -minrate 100k -maxrate 2500k -bufsize:v 1800k -bf 2 -g 15 -c:a libfdk_aac -b:a 128k -ac 2 -f mpegts /mnt/extra/$channel/"$files".conv [16:04:08 CEST] <Admin__> <Admin__> basically its encoding with nvenc [16:04:19 CEST] <Admin__> its complaining about the -tune option not used [16:04:29 CEST] <iive> ripmind: afaik it doesn't have any external dependency, you just need recent version of ffmpeg [16:04:30 CEST] <Admin__> ehh nevermind... [16:04:40 CEST] <Admin__> i changed the preset already my bad [16:04:51 CEST] <ripmind> iive: just Compiled it from git master [16:05:00 CEST] <ripmind> So.. I shoukd have latest [16:05:45 CEST] <ripmind> ffmpeg version N-79998-gfd4a52e [17:44:20 CEST] <Fyr> guys, does FFMPEG support MKV metadata? I tried to set default audio channel - it seems like FFMPEG skips it. [17:44:28 CEST] <Fyr> so as default subtitle [17:45:22 CEST] <Fyr> and comment to metadata, like title. I attached a cover image, titled it, however, the final MKV file doesn't contain the title. [17:48:54 CEST] <Admin__> hey guys... soo strange.. /nvidia_video_sdk_6.0.1/Samples/NvEncoder/NvEncoder -i /dev/stdin -o /dev/stdout -size 1280 720 -bitrate 1800000 -vbvMaxBitrate 2500000 -vbvSize 1000000 -deviceID 0 -preset lowLatencyHQ -qp 21 -numB 1 -goplength 180 -rcmode 32 -fps 30 .. i am using this to encode and the quality is amazing... but when i try to use ffmpeg with v:c nvenc i get horrible quality.. what am i doing wrong ? $ffmpeg_path/ffmpeg -y -i [17:48:55 CEST] <Admin__> "$files" -fflags sortdts -vcodec nvenc -preset llhq -2pass 1 -b:v 1800k -minrate 100k -maxrate 2500k -bufsize:v 1800k -bf 2 -g 15 -c:a libfdk_aac -b:a 128k -ac 2 -f mpegts /mnt/extra/$channel/"$files".conv [17:52:52 CEST] <furq> is the nvenc file actually sticking to the specified bitrate or is it using -qp 21 [17:52:56 CEST] <Aerroon> if i have a video + audio file like a .mkv where the video length is different from the audio length (there's apparently some missing frames at the start of the video file) is it somehow possible to demux said file into a video file and an audio file BUT their lengths would be the same? [17:53:00 CEST] <furq> also the gop length is different [17:53:09 CEST] <Aerroon> so that ffmpeg would not cut out the first few seconds of "no new frames" at the start? [17:57:56 CEST] <Admin__> i tried to use -qp 21 [17:58:12 CEST] <Admin__> oh that... you mean nvtranscoder ? [17:58:23 CEST] <Admin__> it is sticking to 2Mbit/s.. [17:58:24 CEST] <furq> yeah [17:58:30 CEST] <Admin__> and qp is like 21 i thik [17:58:37 CEST] <Admin__> the quality is amazing at the same 1.8Mbit/s [17:58:54 CEST] <Admin__> where as ffmpeg.. it doesn't respect the bit rate with the qmin and qmax at like 21 [17:59:48 CEST] <Admin__> i am a tad confused.. i want to use ffmpeg... with nvenc but .. i can't get over the quality difference. Now why would it respect bit rate and still maintain qp 21 [17:59:59 CEST] <Admin__> i can't seem to do that with ffmpeg [18:00:45 CEST] <Admin__> the problem i am having with nvtranscoder is there must be something up with the decoder because videos get corrupt when i use the input as concat because sizes and audio rates and everything keeps changing. The nvtranscoder can't handle those changes too much i guess [18:01:01 CEST] <Admin__> works very well when source is live stream at mpeg2video .. never changes so the quality is amazing [18:01:36 CEST] <furq> are you sure it's actually 2mbps [18:01:40 CEST] <Admin__> but i am trying to do this with concat files and use nvencoder hardware with llhq and qp 21 [18:01:47 CEST] <Admin__> yes i am sure its 2Mbit/s [18:01:49 CEST] <furq> i'd have thought setting the qp would force it to ignore all the other bitrate settings [18:01:58 CEST] <furq> or to ignore the requested qp [18:01:59 CEST] <Admin__> using vlc , it respects the variable bit rate with min and max [18:02:19 CEST] <Admin__> doesn't seem to do it.. i don't know why to be honest.. but it works very very well [18:02:27 CEST] <furq> shrug [18:02:30 CEST] <furq> can't say i've ever used nvenc [18:02:48 CEST] <furq> the only other obvious difference is the gop length [18:02:54 CEST] <Admin__> yup... and the file sizes.. are hugely different.. its able to compress amazing compared to nvenc with ffmpeg [18:02:57 CEST] <furq> i can't see that making such a huge quality difference [18:03:12 CEST] <Admin__> well that is for testing... gop 180 was the same for ffmpeg .. i must have changed that for testing [18:03:25 CEST] <Admin__> i did try them both at 180 before.. same result [18:04:02 CEST] <furq> why are you passing -2pass 1 [18:04:06 CEST] <Admin__> so, it seems i am back to nvtranscoder unfortunatly :( [18:08:10 CEST] <jkqxz> I don't really believe that bitrate number at all; 2Mbps is not enough for good quality at 720p with any sort of dynamic content (though maybe if your stream has very little activity). Have you verified it by dividing the output file size by the length, rather than looking at possibly-wrong metadata? [18:11:25 CEST] <furq> decent 720 at 2mbit isn't that hard if you have a clean source [18:12:07 CEST] <furq> i'd say "just ask youtube" but it's a stretch to call that decent [18:12:11 CEST] <furq> it's acceptable though [18:12:26 CEST] <furq> granted they're probably not using nvenc [18:13:46 CEST] <jkqxz> Yeah, it can be decent, but frames containing lots of movement will still have visible artifacts. It's not like constant-QP = 21. [18:25:29 CEST] <Admin__> quality is amazing.. even in sports.. i have sports running at 2.2 and it looks great!! [18:25:45 CEST] <Admin__> yes i have a clean source for 720p [18:25:59 CEST] <Admin__> my source is actually 1080p [18:32:08 CEST] <Th3R0b> hi, im wondering if there is an anaglyph output mode for red/blue other then grayscale like there is for red/cyan? [18:32:21 CEST] <Admin__> hey you guys know how i get rid of these errors : [mpegts @ 0x31bcee0] pts (5370840) < dts (5373840) in stream 0 [18:32:21 CEST] <Admin__> [mpegts @ 0x31bcee0] pts (5376840) < dts (5379840) in stream 0 [18:32:21 CEST] <Admin__> [mpegts @ 0x31bcee0] pts (5382840) < dts (5385840) in stream 0 [18:32:33 CEST] <Admin__> is there some filter i can use to clean that up ? [18:36:48 CEST] <Th3R0b> right now im using --vf=lavfi="stereo3d=sbs2l:arbg" but, arbg makes the movie grayscale (black and white), I'd much like something like the dubois (arcb) mode there is for red/cyan... [18:37:13 CEST] <Th3R0b> I suspect im simply overlooking something [18:38:29 CEST] <hmir_> hi guys. I hope this is the riht place to ask this. [18:39:20 CEST] <hmir_> When I run an encoding job, is there some way to decrease the verbosity of the running job? I dn't want to turn it off 'cause then I could just redirect it to /dev/null. i just want to decrease the amount of infomration being spit to screen. [18:43:38 CEST] <rocky_> i need help installed ffmpeg version 2.5.11 on RHEL 6. Is there a yum package for that? [18:51:35 CEST] <c_14> hmir_: you can use '+' to increase and '-' to decrease verbosity [18:56:50 CEST] <Admin__> hey so nobody knows how i can get rid of these errors ? [mpegts @ 0x31bcee0] pts (5376840) < dts (5379840) in stream 0 [19:00:47 CEST] <jkqxz> They indicate that the encoder/muxer making the stream screwed up somehow. Where has the stream come from? [19:03:24 CEST] <tab1293> Does anyone here have any experience using webm_chunks and the webm_dash_manifest formats? [19:04:11 CEST] <tab1293> I get this error when trying to generate the manifest file: Could not write header for output file #0 (incorrect codec parameters ?): Invalid data found when processing input [19:07:13 CEST] <jkqxz> That's the generic "something went wrong" error at the muxer starting. Look further up in the log to see if there are any other errors? [19:10:09 CEST] <tab1293> jkqxz: is there a debug flag to get more info? [19:10:44 CEST] <tab1293> ahh maybe -loglevel debug ? [19:12:50 CEST] <tab1293> hmm doesnt give much more info [19:13:57 CEST] <Admin__> wow.. so NVtranscoder took a 1.9 GIG file.. down to 40Mbit/s and the quality is spectactular [19:14:04 CEST] <Admin__> i can't seem to replicate this with ffmpeg and nvenc [19:14:07 CEST] <Admin__> i am not sure why [19:14:23 CEST] <Admin__> 40MB file.. i mean. not mbits [19:14:33 CEST] <Admin__> 1.9 GIG > 40MB [19:14:47 CEST] <Admin__> at 2Mbit/s with 720 .. and amazing quality.. hard to tell the difference [19:15:09 CEST] <anddam> what library does provide h.265 encoding? [19:15:41 CEST] <JEEB> you can use either libkvazaar or libx265 for it [19:16:32 CEST] <JEEB> kvazaar being newer and not yet as far project that has a community, while libx265 is a MultiCoreWare product that also got released under the GPL because they licensed the name and rights to use code from x264 [19:16:40 CEST] <anddam> I figure I don't have it https://gist.githubusercontent.com/anddam/c35db9804d08f794e651a09839dcb6d8/… [19:16:48 CEST] <anddam> woa long URL, sorry [19:17:11 CEST] <anddam> what's odd is that it doesn't show h.264 that I'm pretty sure I've transcoded [19:17:30 CEST] <JEEB> you can decode it just fine, you only need it for encoding. the only realistic alternative for AVC encoding is libx264 [19:17:44 CEST] <JEEB> and since your configure options don't include --enable-libx264 you don't have it [19:18:04 CEST] <JEEB> both avc and hevc decoders are within libavcodec, it's just encoding that is usually started in a separate project [19:18:15 CEST] <anddam> I see [19:18:46 CEST] <anddam> I have several variants available, libx265 among those [19:19:01 CEST] <anddam> but not h.264 [19:19:03 CEST] <anddam> odd [19:20:08 CEST] <rocky_> i need help installed ffmpeg version 2.5.11 on RHEL 6. Is there a yum package for that? [19:20:08 CEST] <JEEB> x265 is one of those projects where you reall see how much an open source community is beneficial to an encoder. they're making progress but since there's no-one who's got any passion about it they're still having issues trying to optimize the new things added in HEVC. Because you can't just rip those heuristics etc from x264 [19:21:05 CEST] <anddam> JEEB: no I mean I don't see any specific option for enabling x64 [19:21:29 CEST] <JEEB> the configure script has --enable-libx264 [19:21:39 CEST] <JEEB> if that's not enabled you're not building with libx264 linked in [19:21:55 CEST] <JEEB> that is only required for encoding as I said, though. decoding is within libavcodec [19:22:18 CEST] <anddam> I want to encode [19:22:47 CEST] <JEEB> then you need to build with libx264 and you will have to have x264 built first for the configure check to pass [19:23:10 CEST] <anddam> I already have x264 installed, I'm looking at the provided options of the ffmpeg port [19:24:47 CEST] <anddam> does "high10" mean anything to you? [19:24:59 CEST] <furq> yes [19:25:01 CEST] <JEEB> it's a profile [19:25:01 CEST] <anddam> it's a variant that "Build using libx264 with 10-bit output" [19:25:11 CEST] <furq> if you don't know what that is then you don't want it [19:25:22 CEST] <JEEB> basically since libx264 can only be built with 8bit OR 10bit bit depth output [19:25:35 CEST] <anddam> I'm looking at it since it pulls in libx264 [19:25:35 CEST] <JEEB> for general usage you will want 8bit [19:25:52 CEST] <JEEB> because, among other things, 10bit AVC is not HW decode'able [19:25:53 CEST] <anddam> seems that since I was disabling gpl2 in the port building I excluded libx264 [19:26:03 CEST] <JEEB> that makes no sense [19:26:06 CEST] <rocky_> Unrecognized option 'vf' [19:26:07 CEST] <JEEB> wait [19:26:10 CEST] <anddam> well it's a packaging thing [19:26:11 CEST] <JEEB> LGPL vs GPL [19:26:17 CEST] <JEEB> that kind of makes sense [19:26:24 CEST] <JEEB> I thought you disabled gplv3 and that would have made no sense [19:26:28 CEST] <JEEB> because libx264 is GPLv2+ [19:26:29 CEST] <JEEB> but yes [19:26:32 CEST] <furq> by "not hardware decodable" do you mean that all hardware implementers are too lazy to bother [19:26:35 CEST] <anddam> and someone thought it may be clashing, I included a set of "nonfree" codecs [19:26:53 CEST] <anddam> ah here it is--enable-nonfree [19:27:01 CEST] <JEEB> the nonfree ones are generally not too useful with the latest version unless you like HE-AAC [19:27:04 CEST] <furq> you don't need that for x264 [19:27:10 CEST] <JEEB> yeah [19:27:25 CEST] <hmir_> /wc [19:27:42 CEST] <anddam> ohh I see, it was my bad, I thought the nonfree bits clashed with gpl2, but seems they don't, it's just that the resulting binary is not redistributable [19:27:46 CEST] <anddam> but it's for my personal use [19:27:48 CEST] <furq> right [19:28:08 CEST] <anddam> and ofc I just completed the build with libx265 but I now lack x264 [19:28:14 CEST] <anddam> let's rebuild [19:29:36 CEST] <furq> you probably also want to upgrade to a newer version while you're at it [19:29:48 CEST] <anddam> newer than? [19:29:53 CEST] <furq> the one in the paste [19:30:20 CEST] <anddam> I have either @2.8.6 or 20160210 [19:30:28 CEST] <furq> yeah 3.0 is out now [19:30:33 CEST] <anddam> stable? [19:30:33 CEST] <furq> 3.0.2, even [19:30:35 CEST] <furq> yes [19:30:43 CEST] <anddam> I'll file a ticket for upgrading the port [19:31:28 CEST] <anddam> I prefer packaged software to fetching and building by hand [19:35:40 CEST] <anddam> furq: https://github.com/FFmpeg/FFmpeg doesn't have 3.0.2 [19:36:38 CEST] <anddam> all that said I'm not even sure my ipad 2 can decode h.265 [19:45:50 CEST] <Threads> anyway to force ffmpeg to make x264/mkv files constant had in the past variable framerate [19:46:28 CEST] <Threads> i know x264 on videolan has the option of force-cfr does ffmpeg have one ? [19:48:07 CEST] <furq> anddam: that's not an official repo [19:48:12 CEST] <furq> http://ffmpeg.org/download.html#releases [19:48:39 CEST] <Fyr> Threads, what command are you using and what exactly are you willing to get in result? [19:49:23 CEST] <anddam> furq: that was linked from the download page tho' [19:49:58 CEST] <furq> oh so it is [19:50:06 CEST] <furq> i thought it was an unofficial mirror [19:50:22 CEST] <furq> either way it's a mirror [19:50:52 CEST] <furq> it looks to be missing a bunch of releases [20:01:35 CEST] <Threads> Fyr its ok found it force-cfr=1 it is apprently [20:48:38 CEST] <anddam> what's the difference in using aac, libfaac or libfdk_aac? [20:49:28 CEST] <anddam> also I tried -preset help as per https://trac.ffmpeg.org/wiki/Encode/H.264 and I got [20:49:30 CEST] <anddam> x264 [error]: invalid preset 'help' [20:49:32 CEST] <anddam> [libx264 @ 0x7fddf983bc00] Error setting preset/tune help/(null). [20:49:59 CEST] <__jack__> anddam: libfdk_aac is supposed to be better is some cases; however, it requires you to compile you own ffmpeg; aac is built-in, so easier to use; dunno anything about libfaac (almost useless now, maybe) [20:50:47 CEST] <anddam> __jack__: I already built ffmpeg, and libfdk_aac is listed in aac encoding support [20:50:58 CEST] <ChocolateArmpits> __jack__: anddam: devs are saying aac may be better now than libfdk [20:51:12 CEST] <JEEB> for LC-AAC you should no longer require fdk-aac [20:51:16 CEST] <JEEB> but HE-AAC fdk-aac is still the best [20:52:54 CEST] <anddam> I don't really know what those are [20:53:07 CEST] <anddam> my use of video is limited to transcoding in order to watch those on ipad [20:53:32 CEST] <__jack__> lc-aac is the common one, he-aac is the 'bandwidth efficient' one [20:53:35 CEST] <__jack__> anddam: see https://en.wikipedia.org/wiki/High-Efficiency_Advanced_Audio_Coding [20:54:29 CEST] <JEEB> basically you don't want HE-AAC if you value quality and can do >=64kbps for stereo [20:54:51 CEST] <ChocolateArmpits> just use aac, it's as good as any other [20:55:33 CEST] <ChocolateArmpits> it also helps that I don't have to compile libfdk anymore [20:56:14 CEST] <__jack__> ChocolateArmpits: exactly :) [20:56:39 CEST] <__jack__> using stock ffmpeg is so much easier [20:57:56 CEST] <furq> anddam: if you're on 2.8.6 then stick with fdk [20:58:07 CEST] <furq> the improvements to the builtin encoder didn't land until 3.0 [20:58:12 CEST] <furq> faac is just garbage [20:58:46 CEST] <furq> also if you're building from ports then you might as well stick with fdk after upgrade, it's probably a bit better [20:58:51 CEST] <furq> pending a proper listening test [20:59:42 CEST] <anddam> furq: no, I'm now using 20160210 that should be a few days before 3.0 [21:00:44 CEST] <furq> well yeah i still use fdk but i wouldn't go out of my way to add it [21:01:02 CEST] <furq> the builtin encoder is good enough [21:01:44 CEST] <anddam> wow I'm hitting 33 fps at reencoding [21:01:47 CEST] <anddam> old i5 [21:01:52 CEST] <furq> also faac is no longer supported afaik [21:02:12 CEST] <furq> it was only supported because there was no better gpl-compatible aac encoder [21:02:59 CEST] <JEEB> which faac ended up not even being [21:03:04 CEST] <JEEB> which is why it's nonfree [21:03:25 CEST] <furq> oh [21:03:29 CEST] <furq> am i thinking of vo-aacenc [21:04:41 CEST] <JEEB> yeah, that one was apache v2 [21:04:45 CEST] <JEEB> and actually licensed as such [21:05:02 CEST] <JEEB> faac was like "wait this code is suspiciously similar to the reference one which is not GPL compatible" [21:05:40 CEST] <anddam> can code not be compatible with a license? [21:05:58 CEST] <anddam> I thought the compatibility applied among licenses [21:06:04 CEST] <__jack__> anddam: nope [21:06:19 CEST] <__jack__> anddam: licence can forbid redistribution of binaries, however [21:06:33 CEST] <JEEB> if you have one license that limits you more than the other, and the less limiting one says you can't apply more limitations then those two licenses are not compatible [21:07:23 CEST] <JEEB> f.ex. fdk-aac has a custom license that says you can't distribute it under X or Y, which constitutes extra limitations. Thus it is incompatible with GPL and most probably LGPL - although IANAL. But that is why fdk-aac requires enable-nonfree [21:07:54 CEST] <JustinHitla> I use that command to concatenate 2 videos 1.avi and 2.avi: "ffmpeg -f concat -i list.txt -c copy -y -f avi out.avi", and these are ffprobe output for videos http://sprunge.us/cYbM, they both the same format and the same fps, lets assume there are 60 frames in first video and 60 frames, and the thing is in out video there are not 120 frames but [21:07:55 CEST] <JEEB> in case of faac it was copying code from a thing which was under extra limitations, which actually made it incompatible with its own license [21:07:59 CEST] <JustinHitla> 124, and I checked them and there are first 60 frames from first video, as it should be but then it goes like: 61,62,63 frame are all duplicates of 1st frame from second video and 64,65,66 frames are all duplicates of 2nd frame from 2 video, so if I call video 1.avi "a" and video 2.avi "b" it goes like a58,a59,a60,b1,b1,b1,b2,b2,b2,b3,b4,b5 and [21:08:03 CEST] <JustinHitla> instead of normal a58,a59,a60,b1,b2,b3,b4,b5 anyone knows what the reason ? [21:08:43 CEST] <JustinHitla> and of course when I play out video it stutters on place of concatenation because of those duplicate frames [21:09:09 CEST] <JustinHitla> now I even thinking of making video into png pictures and remove duplicates by hand and then reassemble them again [21:10:05 CEST] <JustinHitla> I even concatenated videos without audio track, thinking if that could be encoder adjasting for async audio video, but that doesn't helped also [21:10:35 CEST] <furq> anddam: https://github.com/mstorsjo/fdk-aac/blob/master/NOTICE#L57-L58 [21:10:39 CEST] <furq> in case you're wondering what all the fuss is about [21:10:42 CEST] <JustinHitla> if nothing, I really go that video->png->video way [21:11:35 CEST] <JustinHitla> anyway, is my question even clear ? [21:12:04 CEST] <JustinHitla> so other words, concatenation made 4 duplicate frames appear in output video [21:14:16 CEST] <JustinHitla> I want to say I concatenated many videos and sometimes it just happens like that, have no idea what is the reason [21:17:17 CEST] <JustinHitla> also first piece I got assembling png files using "ffmpeg -f image2pipe -vcodec png -r 29.970" command and second piece I got cutting out it from other video using that command "ffmpeg -ss 27 -i _/2fyzySVrA7A.mp4 -c:v mpeg4 -vb 10000k" but that shouldn't be a problem [21:28:28 CEST] <arbi> I have an interlaced 720x576 AVI file but for some reason Ubuntu sees it as 720x288 (ffmpeg reads it correctly). Can I somehow correct this information without re-encoding the whole thing? [21:28:31 CEST] <JustinHitla> herp derp [21:29:39 CEST] <JustinHitla> Ubuntu sees it ? [21:29:48 CEST] <JustinHitla> how Ubuntu a video software ? [21:30:15 CEST] <arbi> JustinHitla: When I look at file properties [21:30:34 CEST] <JustinHitla> that is file manager [21:30:48 CEST] <arbi> Also when I play with Videos (the software) [21:31:03 CEST] <arbi> JustinHitla: Yes it's the file manager [21:31:49 CEST] <arbi> Video editing program Cinelerra also decodes it as 720x288.. half the screen becomes green :-/ [21:32:45 CEST] <arbi> Have you seen this before? [21:32:55 CEST] <JustinHitla> no [21:37:07 CEST] <arbi> eh I'll re-encode it [21:43:27 CEST] <JustinHitla> arbi: do you know by any chance why I got duplicate frames when concatenated 2 videos ? so now output video stutters in place of concatenation because of those frames [21:44:29 CEST] <arbi> I've only done that with Cinelerra [22:06:57 CEST] <JustinHitla> herp derp [22:11:44 CEST] <_Vi> Is update of FFmpeg to version 3.0 more breaking than usual? Has the major version bumped per semantic versioning or just to avoid two-digit numbers in minor version? [22:12:22 CEST] <JEEB> there can always be major updates in non-step releases [22:12:37 CEST] <JEEB> so in that sense "3.0" is no different [22:13:03 CEST] <JEEB> the libraries all have separate versioning so semantic versioning really doesn't make too much sense in the project as a whole [22:13:14 CEST] <JEEB> the libraries do try to abide to semantic versioning [22:14:04 CEST] <_Vi> Semantic versioning should tell me if my scripts employing complicated FFmpeg filter chains guarantied not to break or not. [22:14:26 CEST] <JEEB> that would be libavfilter version then [22:14:57 CEST] <JEEB> it's simpler when a thing has a single library around :) [22:15:18 CEST] <JEEB> in this case there's libavcodec, libavformat, libavfilter, libswscale, libswresample, libavresample, libavutil... [22:15:28 CEST] <_Vi> I.e. bump major version number on removal of some effect, renaming options (without leaving deprecated variant), changing how command line is parsed i general. Middle version bump - new effects, new formats, new options. Patchlevel bump - fixes, optimisations. [22:15:58 CEST] <_Vi> I mean the version of ffmpeg command line tool (maybe also ffserver and ffprobe). [22:16:11 CEST] <JEEB> yeah, I've generally seen that on API side, not so much on cli side. Generally ffmpeg.c cli breakage is frowned upon, of course [22:17:18 CEST] <_Vi> Bumping major version may hint that CLI breakage has finally occurred (e.g. major CLI refactor) and hints packaging FFmpeg to allow co-existance of FFmpeg2 and FFmpeg3. [22:17:48 CEST] <_Vi> So FFmpeg 3.0 is basically FFmpeg 2.9. [22:17:50 CEST] <JEEB> yes, but the only thing with FFmpeg currently is that you get point releases for releases [22:18:04 CEST] <JEEB> and otherwise the main release version can be whatever the person making the release fancies [22:18:32 CEST] <JEEB> for cli generally testing a new release should be simple enough since you can build a static FFmpeg and test that once [22:18:45 CEST] <JEEB> only after that you can actually start migrating to it [22:18:54 CEST] <_Vi> I think semantic versioning should also be used for command line tools scripts may rely upon. [22:19:05 CEST] <_Vi> CLI tools are "libraries" for Bash. [22:19:25 CEST] <JEEB> I don't disagree, I just don't see too much care about releases in general here :P they're there mostly for distros that really like having versions [22:19:32 CEST] <__jack__> that's what people do [22:19:44 CEST] <JEEB> also as I said, ffmpeg.c usage breakage is generally frowned upon [22:20:05 CEST] <JEEB> like, we have the -c:v et al syntax, but -vcodec et al are not yet removed and probably will never be [22:21:04 CEST] <JEEB> it does happen every now and then of course, like the removal of the ultra-crappy "deinterlacer" which wasn't really a deinterlacer (-deinterlace option I think?) [22:21:44 CEST] <_Vi> I'd prefered FFmpeg's major version remaining "2" unless there is finally a decision of remove all those deprecated things and/or change how CLI works in general. [22:22:27 CEST] <_Vi> As "3.0" already happened, maybe it should be called "3.00" to prevent further needless "overflowing" to "4.0"? [22:22:44 CEST] <JEEB> you could make a proposal about that, but the issue is that ffmpeg.c usage compatibility is generally broken so rarely... :) [22:23:23 CEST] <_Vi> JEEB, A library whose API is never broken can be ever v1.* and this is a good thing. [22:24:17 CEST] <JEEB> anyways, me or anyone I'm related to isn't using the releases so while I do understand and more or less agree with your point... I don't have any strong feelings any way :) [22:24:39 CEST] <_Vi> (I also don't like how Linux kernel version is managed. One cannot specify "I want >= 4.4.8, < 5.0.0" and expect things not to break). [22:24:41 CEST] <JEEB> (and not to mention I'm not someone who can do something about it without doing something proactive about it) [22:25:13 CEST] <JEEB> (and because I don't really gain much from any versioning scheme changes I'm going to be a lazy git) [22:25:31 CEST] <JEEB> but as I said, feel free to propose it on trac or the mailing list [22:25:39 CEST] <_Vi> Maybe somebody has specified in some dependency list "FFmpeg > 2.7, < 3.0" and there are already missed upgrade opportunities because of 3.0-ness. [22:26:19 CEST] <JEEB> well if it's a dependency that's a bigger issue, and that person probably wouldn't be so liberal in general [22:26:43 CEST] <JEEB> because APIs can and do change between each (branched-off) release [22:27:23 CEST] <JEEB> so if a person knows that their thing would work with 2.7.x and 2.8.x they most probably wouldn't automagically let it grab >2.8 [22:27:39 CEST] <_Vi> JEEB, You have told that FFmpeg's API (that is CLI) should not change. And FFmpeg's libraries have their own version numbers. [22:27:49 CEST] <JEEB> yes [22:28:17 CEST] <JEEB> basically it depends on if you're packaging the whole of ffmpeg under "ffmpeg" or just the cli, although the cli depends on the libraries [22:28:27 CEST] <JEEB> and if you're dependant on the cli or the libraries [22:28:59 CEST] <JEEB> yes, with libraries you usually have stuff like libavcodec > 56 [22:29:33 CEST] <_Vi> Sometimes I write programs that use FFmpeg as a subprocess, without linking to the libraries. Obviously I need to declare dependency on FFmpeg CLI tool. What version range should I use? [22:32:28 CEST] <JEEB> generally given that ffmpeg.c cli usage isn't usually broken, you set it to ><the lowest version you've tested to work> [22:32:50 CEST] <JEEB> and then try to test master periodically, that should keep you up to snuff [22:34:11 CEST] <JEEB> general health of FFmpeg can be seen with http://fatebeta.ffmpeg.org/ [22:36:06 CEST] <_Vi> So you recommend just lower version bound. OK. [22:36:45 CEST] <JEEB> yes, but that does require semi-constant quick checks with latest stuff. just in case a breakage occurs for you. it is, as I said, rather rare though [22:37:34 CEST] <JEEB> most projects depending on FFmpeg do try to follow the master relatively closely so they spot any breakage before it gets pushed out into a release [22:38:56 CEST] <MrSassyPants> How do you enable scene cut keyframes in libvpx / vp9 in ffmpeg? I can't seem to find it in the documentation [22:39:56 CEST] <JEEB> libvpx unfortunately still sucks for my use cases so E_NO_IDEA from me. But I would have guessed that'd be the default for any sane encoder :P [22:40:13 CEST] <JEEB> -g would set the maximum GOP length, and then IRAPs would be set according to the clip [22:40:47 CEST] <JEEB> of course I wouldn't be surprised if libvpx's scenechange stuff wouldn't be exactly too great (although I'd think that'd be one of the simpler things to get done) [22:42:18 CEST] <_Vi> The point of semantic versioning is to avoid such checks. That looks as if there is invisible major version number (like FFmpeg "1.3.0.2" where "1." is hidden, invisible version number) with a promise that this invisible major version number will never bump. But will this "never" last, for example, 10 years? [22:44:32 CEST] <JustinHitla> "invisible monsters, thats going to be fun" -- Barry [22:46:16 CEST] <JEEB> _Vi: I do understand what you're aiming at but unfortunately that's not how it works :) and when stuff is decided to be broken it will be - generally by outputting a warning and then plugging it out. [23:00:19 CEST] <JustinHitla> are there any specs my videos must conform to in order to be concatenated properly ? [23:00:35 CEST] <JustinHitla> should I convert both to some generic format and then concatenate ? [23:00:53 CEST] <JustinHitla> I think it concatenates on "stream" level not file level so it shouldn't be an issue [23:13:47 CEST] <JustinHitla> I got "[mpeg4 @ 0x8190a40] warning: first frame is no keyframe" could it be when I concatenate two videos and in second one first frame is not a key frame I get those duplicate frames ? [00:00:00 CEST] --- Mon May 16 2016

1 0

ffmpeg-devel.log.20160514
by burek 15 May '16

15 May '16

[00:08:41 CEST] <KGB> [13FFV1] 15michaelni pushed 1 new commit to 06master: 02https://git.io/vr3wc [00:08:41 CEST] <KGB> 13FFV1/06master 148a60355 15Jérôme Martinez: Rewrite of the restrictions paragraph... [02:41:39 CEST] <Illya> where is AVFormatContext->codec->sample_rate now? (as it's depreciated) And, is there somewhere where I can see the deprecations in a centralised way? [02:47:21 CEST] <Illya> Ah, it's in codecpar. [04:00:12 CEST] <cone-818> ffmpeg 03James Almer 07master:004b5823860d: avutil/lzo: split test out into a separate file [04:48:57 CEST] <cone-818> ffmpeg 03ZhouXiaoyong 07master:d680ab1c46b6: avutil/mips: header asmdefs.h add some PTR_ macros for loongson [14:31:37 CEST] <cone-728> ffmpeg 03Paul B Mahol 07master:51a1c1c6ac98: avcodec/iff: check for end of buffer in input for delta_l decompression [20:02:50 CEST] <cone-345> ffmpeg 03foo86 07master:365b0c13e461: avcodec/dca: print error message when no frame is found in packet [00:00:00 CEST] --- Sun May 15 2016

1 0

ffmpeg.log.20160514
by burek 15 May '16

15 May '16

[12:45:26 CEST] <Amitari> I just found out that ffmpeg has added support for IFF-ANIM, which is awesome. I tried to convert an ANIM to GIF, with "ffmpeg -i file.anim file.gif". It worked, however, it wasn't animated. Does anyone know how to use this properly? [12:49:50 CEST] <durandal_1707> Amitari: old ffmpeg, need master [12:51:35 CEST] <Amitari> Oh, so earlier versions just had some single-frame conversion of IFF? [12:54:03 CEST] <iive> durandal_1707: maybe we should add a check of how old is the ffmpeg version. [12:54:55 CEST] <durandal_1707> iive: sorry can't parse [12:55:02 CEST] <iive> e.g. after printing the copyright notice, we can print red color message "Your ffmpeg is 1 year old". [12:55:42 CEST] <furq> can we also license ffmpeg under the wtfpl and then throw a fit when a distro tries to remove the message [12:56:38 CEST] <iive> Is wtfpl a thing? It should be! [12:56:55 CEST] <furq> yes it is [12:56:57 CEST] <furq> please don't use it [12:59:06 CEST] <durandal_1707> why? It's best one. [13:00:28 CEST] <furq> it's not a real license and will more than likely prevent companies from using your software [13:01:03 CEST] <Mavrik> Or anyone with a bit of brain :P [13:01:08 CEST] <iive> no warranty disclaimer. [13:01:19 CEST] <furq> also i've seen at least three instances now of someone throwing a fit because someone did what the fuck they wanted to with wtfpl-licensed code [13:01:43 CEST] <furq> actually two, because one was just baseless accusasions of plagiarism of wtfpl-licensed code [13:02:52 CEST] <furq> er. accusations [13:12:14 CEST] <durandal_1707> name it? [13:15:09 CEST] <furq> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=819703 [13:15:21 CEST] <furq> that was the most high-profile one [16:40:25 CEST] <anjum> I cloned and built ffmpeg on Debian, but I cannot use the VPx codec (for converting to webm). I downloaded and extracted libvpx-1.5.0.tar.bz2 to a folder in my sources directory, tried to run configure --enable-libvpx, but it warns 'libvpx decoder version must be >=0.9.' [16:41:34 CEST] <__jack__> anjum: simple solution: install from Debian repositories [16:43:31 CEST] <anjum> I will try that again. [16:46:00 CEST] <anjum> I tried to do apt-get install ffmpeg earlier, but it couldn't find the package. [16:56:20 CEST] <__jack__> anjum: use the unstable repo [17:02:01 CEST] <anjum> It's installing now, seems promising. [17:07:58 CEST] <anjum> It works, thank you. [17:46:10 CEST] <Zaoshi> Hello. I'm trying to extract image from .mkv file. When I use "ffmpeg -y -ss TIME -i MY_FILE -copyts -vframes 1 OUT_FILE" it's instant, but if I add subtitles ("ffmpeg -y -ss TIME -i MY_FILE -copyts -vf "subtitles=MY_FILE" -vframes 1 OUT_FILE") it takes a very long time to generate image. Is there any way to make it faster? [18:05:36 CEST] <Zaoshi> Hm... Apparently extracting subtitles before hand makes it instant as well. Why does it take so much more time to read subtitles while rending image than it takes to extract subtitles + render image separately? [18:38:06 CEST] <thebombzen> Zaoshi: the subtitles filter doesn't work well with -ss [18:39:05 CEST] <thebombzen> my best guess is that when you started using subtitles, it went to slow decoding. [18:39:36 CEST] <thebombzen> my recommendation for a single frame is to extract the frame and just put the subs on manually with something like GIMP unless they're fancy or complicated [18:39:54 CEST] <thebombzen> because of the problem you just stated. [18:40:49 CEST] <thebombzen> I mean, in an ideal would you wouldn't have to do ffmpeg -i input.mkv -vf subtitles=video.mkv. In an ideal world you could do that with -filter_complex, i.e. ffmpeg -i input.mkv -filter_complex rendersubtitles or something like that. [18:40:59 CEST] <thebombzen> Unforutunately FFmpeg doesn't support that yet [19:02:01 CEST] <Zaoshi> Alright. Thanks for the info. For now I'll just extract subtitles into separate file and use it as 2nd source. Solves the issue but takes a bit time to extract subtitles. [21:18:50 CEST] <ElAngelo> if you compare ffmpeg today with ffmpeg 2 years ago... what would be the greatest changes/improvements? [21:31:04 CEST] <klaxa> aac encoder? [21:33:05 CEST] <kid195> afternoon all [21:35:30 CEST] <kid195> anybody home ? [21:43:24 CEST] <durandal_1707> disco party [21:50:00 CEST] <DHE> this Alf marathon isn't going to watch itself [21:54:03 CEST] <ElAngelo> any of you know how one would proceed to use ffmpeg and be completely save legally? [21:54:08 CEST] <ElAngelo> safe [21:54:32 CEST] <klaxa> you use it [21:54:37 CEST] <klaxa> and you don't go to jail [21:54:51 CEST] <klaxa> if you want to ask licensing questions you need to be a lot more specific [21:57:53 CEST] <kid195> anybody know how I can play videos like live tv using ffmpeg with red5 ? [21:58:17 CEST] <kid195> ffmpeg -re -i /usr/local/red5/webapps/oflaDemo/streams/Avengers2.mp4 -f mp4 rtmp://localhost/oflaDemo/livestream1 i was working with that for a test, but no luck [21:58:37 CEST] <kid195> i got input/output error [21:58:50 CEST] <ElAngelo> nevermind just read te whole licensing and legal thingy [21:58:54 CEST] <ElAngelo> it's quite clear :) [22:04:51 CEST] <kid195> anyone here ?? [22:27:45 CEST] <twak54> hi [23:18:02 CEST] <tab1293> So I am trying to use the webm_dash_manifest format to generate a mpd file for some webm_chunks [23:20:50 CEST] <tab1293> anyone @here? [00:00:00 CEST] --- Sun May 15 2016

1 0

ffmpeg-devel.log.20160513
by burek 14 May '16

14 May '16

[00:05:43 CEST] <iive> pfelt: better reply in the maillist... [00:06:28 CEST] <pfelt> i'm resubmitting now [00:07:51 CEST] Action: pfelt is an idiot [00:07:52 CEST] <pfelt> :( [00:08:25 CEST] <wm4> everyone makes mistakes, and that was a minor one [00:09:13 CEST] <pfelt> for some stupid reason i assumed that av_usleep was defined in a header already included&.. and i neglected to compile&. lesson learned [00:13:17 CEST] <rcombs> I've sent much worse breakage to the ML [00:13:50 CEST] <rcombs> (this is why we C/R) [00:27:05 CEST] <cone-256> ffmpeg 03Paul B Mahol 07master:cb928fc448f9: lavc: add IFF ANIM decoder [00:42:17 CEST] <pfelt> so i asked this in the main channel, but maybe someone here knows. do we have any output formats that are multithreaded? decklink is single thread and i *think* that's why i can't do && -f decklink 'somethin' & -f decklink 'somethinelse' and have it perform well [03:53:08 CEST] <cone-230> ffmpeg 03Tobias Rapp 07master:77d1e88cf528: avformat/dump.c: fix mixed log levels [03:53:08 CEST] <cone-230> ffmpeg 03Andriy Lysnevych 07master:ed3a02547c71: avcodec/avpacket: Respect payload offset in av_packet_ref [06:07:46 CEST] <Zeranoe> Interesting (albeit subjective) writeup on the internal AAC encoder https://hydrogenaud.io/index.php?topic=111085 [07:29:21 CEST] <rcombs> > it was latest at the time but TNS algorithm was fixed since then [11:59:21 CEST] <j-b> is LBR a new CODEC_ID? [12:02:47 CEST] <nevcairiel> no [12:06:51 CEST] <j-b> nevcairiel: that's weird, but OK. [12:07:09 CEST] <j-b> nevcairiel: thx [12:20:17 CEST] <jkqxz> wm4: Hmm, I didn't immediately see how nasty that was. It would be best if it were unreffed in avcodec_free_context(), and not touched in avcodec_close()? (Like rc_override and others there.) [12:20:48 CEST] <nevcairiel> having something user allocated and then taken over by avcodec seems generally a bit iffy [12:22:02 CEST] <wm4> let's ask elenril! [12:27:35 CEST] <wm4> nice, scam spam [14:03:50 CEST] <cone-182> ffmpeg 03Paul B Mahol 07master:17ea1cd55960: avcodec/iff: change delta l offset to uint32_t [19:19:29 CEST] <jamrial> ubitux: your fate clients haven't updated in four days [21:31:34 CEST] <cone-691> ffmpeg 03Matthieu Bouron 07master:f6265a5cbcfb: swresample/arm: add ff_resample_common_apply_filter_{x4,x8}_{float,s16}_neon [21:56:57 CEST] <cone-691> ffmpeg 03James Almer 07master:f92d2f329254: avutil/aes-test: fix memleak [21:56:58 CEST] <cone-691> ffmpeg 03James Almer 07master:64dbfcc415c5: avutil/ripemd-test: fix memleak [21:56:58 CEST] <cone-691> ffmpeg 03James Almer 07master:7ee0e0ca1d51: avutil/sha-test: fix memleak [21:56:59 CEST] <cone-691> ffmpeg 03James Almer 07master:d8ef8b960add: avutil/sha512-test: fix memleak [22:02:11 CEST] <jamrial> BBB: http://pastebin.com/TKwZGey8 memleak in you matroska-remux test [22:02:22 CEST] <BBB> :( [22:02:30 CEST] <jamrial> although it may not be your code at fault [22:05:05 CEST] <BBB> its a leak in the general code [22:05:09 CEST] <BBB> the compat layer between old and new bsfs [22:07:49 CEST] Action: jamrial pokes rcombs [22:10:29 CEST] <BBB> it seems like the auto-bsf are never free'ed [22:10:43 CEST] <BBB> I dont know whose responsibility that would be (muxer implementation or generic code) [22:10:47 CEST] <BBB> rcombs: you should know that :D [22:12:06 CEST] <BBB> did I write the first test that relies on auto-bsf? [22:12:13 CEST] <BBB> Im so proud of myself, I found an actual bug ;) [22:12:42 CEST] <jamrial> technically the first to be commited. rcombs has like three in his auto-bsf patchset :p [22:29:30 CEST] <cone-691> ffmpeg 03James Almer 07master:8ba257bac642: avformat/Makefile: fix avlanguage dependencies [22:38:14 CEST] <BBB> jamrial: Ill take whatever I can get ;) [22:41:19 CEST] <rcombs> BBB: jamrial: that's fixed in my patchset [22:41:26 CEST] <rcombs> (the leak, that is) [22:41:28 CEST] <BBB> cool [23:28:44 CEST] <cone-691> ffmpeg 03Paul B Mahol 07master:01ead2a63963: avcodec/iff: implement 3rd delta ANIM decompression [00:00:00 CEST] --- Sat May 14 2016

1 0

ffmpeg.log.20160513
by burek 14 May '16

14 May '16

[01:27:32 CEST] <voltagex> this is a silly question probably, but will an i7 2600k always outperform an i7 6560U in encoding performance with libx265 due to higher clock speed? I'm seeing a 30-50% encoding speed increase in favour of the 2600k. Guess Skylake doesn't help that much? [01:39:29 CEST] <furq> voltagex: it's more down to TDP than clock speed [01:39:47 CEST] <ploc> hi I'm using the following ffmpeg command : ffmpeg -i "${inputVideo}" \ -c:a copy \ -c:v libx265 \ -crf 28 \ -preset:v medium \ -profile:v main \ "${outputVideo}" [01:40:05 CEST] <ploc> and I get an error : [libx265 @ 0x942e1e0] [Eval @ 0xbfa18388] Undefined constant or missing '(' in 'main' [libx265 @ 0x942e1e0] Unable to parse option value "main" [libx265 @ 0x942e1e0] Error setting option profile to value main. [01:40:37 CEST] <ploc> I'm using ffmpeg version 2.8.6-1~bpo8+1 [01:47:37 CEST] <Rudde> Hey [01:47:41 CEST] <llogan> ploc: there is no -profile private option for libx265 in ffmpeg [01:48:12 CEST] <llogan> and i doubt it will work via -x265-params: http://x265.readthedocs.io/en/default/cli.html#profile-level-tier [01:48:27 CEST] <Rudde> Is it possible to pass multipler values to -ss for points to make screenshots? [01:49:15 CEST] <ploc> llogan: ok, it is strange because I've just read it in a french magazin which is usually reliable [01:49:50 CEST] <llogan> you can use --profile is you use x265 binary directly [01:50:09 CEST] <llogan> but just try without it. it will probably just choose main anyway [01:50:35 CEST] <llogan> Rudde: probably easier to use select filter [01:51:43 CEST] <Rudde> I would like to to be fast and just get 3, 4 or 5 fullscreen uncompressed png screenshots from a video want it to be very fast and hopefully meanyfull screenshots? And my experience it that select is very slow? [01:52:07 CEST] <llogan> -vf "select='eq(t,0)+eq(t,25)+eq(t,50)+eq(t,75)'" -vsync vfr output_%03d.png [01:52:15 CEST] <llogan> yes, it is probably slower than -ss [01:52:46 CEST] <ploc> llogan: you're right it does work without the profile parameter (maybe the guy is using libav???) [01:53:46 CEST] <Rudde> Would I be better of by just running with -ss 3-5 times? [01:54:09 CEST] <llogan> ploc: you can remove preset too since medium is the default [01:54:26 CEST] <llogan> Rudde: i don't know. compare both methods and time it [01:55:07 CEST] <Rudde> seems like select us extrmly slow, even with video in ram [01:57:23 CEST] <furq> `-vf select='gt(scene\,0.4)' -frames:v 4 out%02d.png` will probably do a better job of getting "meaningful" screenshots [01:57:29 CEST] <furq> it'll still be slow though [01:59:35 CEST] <ploc> llogan: according to your link, profiles are supported in h265: The following profiles are supported in x265. [01:59:38 CEST] <Rudde> That gave me 4 black screens :P [01:59:44 CEST] <Rudde> where the first one gave me only one [01:59:52 CEST] <Rudde> but all seemed to be from the very beginning of the clip [02:01:32 CEST] <llogan> ploc: yes, but that option says "CLI ONLY: API users must blah blah blah". [02:02:11 CEST] <ploc> llogan: how do you use ffmpeg without CLI? [02:03:14 CEST] <llogan> it's referring to x265 cli [02:09:40 CEST] <voltagex> furq: any more information? Do you mean thermal throttling? [02:11:05 CEST] <furq> i mean the fact that one is a laptop cpu and one is a desktop cpu [02:11:41 CEST] <furq> the clock speed and architecture don't really mean anything when you're comparing an 85 watt cpu with a 15 watt one [02:33:42 CEST] <Rudde> What is the difference between -vframes and -frames? [08:30:49 CEST] <LigH> Hi. [08:33:02 CEST] <LigH> According to Wikipedia, libavcodec shall support decoding CineForm HD (GoPro). But in the full help of an ffmpeg recently built with a suite in "non-free" mode, claiming to support all available codecs, I could not find it mentioned as supported format. Does it require a specific build switch they may have forgotten to include? [08:37:42 CEST] <relaxed> LigH: ffmpeg -codecs 2>&1| grep Cineform [08:40:58 CEST] <furq> LigH: anything newer than 3.0 should have it [08:42:25 CEST] <furq> what a nice young man [08:42:52 CEST] <LigH> Sorry, our ISP has technical issues. This is a "No-DSL area". [08:42:59 CEST] <furq> 07:40:58 ( furq) LigH: anything newer than 3.0 should have it [08:43:44 CEST] <LigH> ffmpeg version N-79976-g34a5e64-g97946b2+1 [08:44:06 CEST] <LigH> auto-built yesterday [08:44:08 CEST] <relaxed> LigH: did the command I gave you return anything? [08:44:37 CEST] <relaxed> if you have support it will return " D.V.L. cfhd Cineform HD" [08:44:54 CEST] <LigH> Yes, that's included. [08:45:17 CEST] <relaxed> then you're good to go [08:45:34 CEST] <LigH> So the codecs list is not included in the full help... [08:45:46 CEST] <relaxed> correct [08:46:11 CEST] <relaxed> well, maybe cfhd is (the name of the decoder) [08:46:15 CEST] <furq> it isn't [08:46:42 CEST] <relaxed> ffmpeg -h decoder=cfhd [08:46:43 CEST] <furq> "-codecs" is mentioned in full help though [08:47:01 CEST] <furq> so it's pretty much your fault for not reading all 8405 lines of -h full [08:47:22 CEST] <furq> or however many your build has. probably more if it supports "all available codecs" [08:49:31 CEST] <LigH> If I want "full help", somehow I expected that the result of "-codecs" was included there as well... [08:50:59 CEST] <relaxed> "-decoders show available decoders" is near the top of ffmpeg -h [08:51:50 CEST] <LigH> I expect all this to be included in "-help full", or this doesn't really mean "full". ;) [08:54:00 CEST] <LigH> But okay. CFHD should be supported for decoding. So it may also be available in related projects, like LAV FIlters. [08:54:10 CEST] <LigH> Provided the ffmpeg base is current. [09:02:00 CEST] <LigH> \o [13:18:53 CEST] <Dariush> Hi. I want to crop some videos, but I don't want to reencode them due to size. Is it possible to set a viewport or something so that the unwanted parts are not shown? [13:20:43 CEST] <JEEB> mp4 has stuff like that but I'm pretty sure very few things actually support it [13:20:58 CEST] <JEEB> matroska kind of has stuff like that, but that is a feature that is very badly defined [13:21:07 CEST] <JEEB> so you should refrain from using it [13:23:42 CEST] <Dariush> oh, all right. thanks ;) [14:28:34 CEST] <gml> Hello, I'm trying to stream the video from my webcam to IIS's Live Smooth Streaming. When I open the stream using VLC, the video starts playing from the moment I executed the ffmpeg command. What I'm trying to achieve is to make clients see the playback from the moment they connect to the stream, not from the very beginning. How can I achieve this? This is my command and its initial output: http://pastebin.com/nj65AMSj [15:57:22 CEST] <thunfisch> hey, trying to capture with decklink module and 10.6.4 driver at 720p50 over hdmi on a decklink mini recorder - only getting color bars. source is 100% setup correctly. anyone else have this problem? [18:09:29 CEST] <nMaib0> why can't libopus change the sample rate to 44.1 KHz? [18:13:28 CEST] <vade> functionally, is there any difference between using libswresample or an avfilterchain with aresample=N and aformat=FMT ? [18:17:19 CEST] <JEEB> I think aresample/aformat internally use swresample [18:17:27 CEST] <JEEB> or at least they should be using that [20:46:09 CEST] <dbugger> Hello guys and girls [20:50:44 CEST] <dbugger> I have a question: I get a video as input with an unknwon resolution. I use this for cropping: "[0:v]crop=w=320:h=240[v0c]". The problem here is that I would like to resize first, so that either the width or the match one of those 2 coordinates, and the crop the rest. The problem is that I cant know before hand if the video will be taller or wider... how could I do this? [20:51:55 CEST] <DHE> there is some formula support in ffmpeg. you could write some very basic logic for it [20:55:37 CEST] <dbugger> I have no idea how to do that... [21:06:03 CEST] <xintox> is it possible to tell ffmpeg to pick a specific stream from a manifest file? [21:06:18 CEST] <xintox> i have to use the manifest file because it sets cookies [21:06:35 CEST] <xintox> using m3u8 [21:14:43 CEST] <klaxa> dbugger: maybe see: https://www.ffmpeg.org/ffmpeg-filters.html#scale-1 ? [21:14:53 CEST] <klaxa> oh wait... [21:14:58 CEST] <klaxa> with cropping... that's a tough one [21:25:22 CEST] <dbugger> klaxa, yeah... [21:26:12 CEST] <klaxa> if the file is not "streamed" you could read it first with ffprobe (or ffmpeg -i) extract the video's dimensions, do some calculations with those and use these as parameters for ffmpeg [21:36:53 CEST] <dbugger> Already working on a solution... wait [21:47:57 CEST] <dbugger> Great, found it :) [21:48:00 CEST] <dbugger> Thanks all anyway :) [22:52:31 CEST] <midgetspy> ffmpeg -ss 600 -i test.mkv -y -vframes 1 -an out.png gives me http://imgur.com/ovFU5iB but if I watch the video myself I don't see any such corruption. is there an option I'm missing to fix this? [23:39:01 CEST] <SleepyPikachu> I'm trying to create a photo slideshow from *.jpg where each image shows for 10s on a loop and the video is as long as the back track audio. I can't figure out how to make image show for 10s [23:39:57 CEST] <SleepyPikachu> This is what I've tried so far: ffmpeg -loop 1 -pattern_type glob -i '*.jpg' -i back.mp3 -c:v libx264 -c:a copy -shortest out.mp4 [00:00:00 CEST] --- Sat May 14 2016

1 0

ffmpeg-devel.log.20160512
by burek 13 May '16

13 May '16

[00:48:54 CEST] <cone-199> ffmpeg 03James Almer 07master:7d8e19a5c4d2: avutil: make crypto testprogs include headers only [03:50:30 CEST] <cone-303> ffmpeg 03Michael Niedermayer 07master:65ffc0b1ed04: avutil/float_dsp-test: Add include config.h for HAVE_* [03:50:30 CEST] <cone-303> ffmpeg 03James Almer 07master:cd244fae9847: avutil/cpu-test: Fix includes (needed for HAVE_*) [09:08:56 CEST] <cone-646> ffmpeg 03Carl Eugen Hoyos 07master:d1cacbbea946: lavc/libutvideoenc: Cast an unsigned constant to int. [09:10:51 CEST] <esdwdftty> http://www.streamingmedia.com/Articles/ReadArticle.aspx?ArticleID=110383 [09:11:02 CEST] <esdwdftty> https://aomedia.googlesource.com/aom/ [09:11:32 CEST] <esdwdftty> !!!!!!!!!!!!!!!!!!!!!!!!!! [09:18:16 CEST] <esdwdftty> 3421 [09:48:45 CEST] <fritsch> wm4: ping [09:49:02 CEST] <fritsch> wm4: http://forum.kodi.tv/showthread.php?tid=231955&pid=2332729#pid2332729 <- did you ever check how exact vaapi's hw decoder works? [09:50:13 CEST] <fritsch> there is a guy on our forum that used an hdmi capture card and it seems that vaapi supresses some details in its output. sw decoding is fine. Currently let him try with mpv to see if it has the same issue. The intel people most likely want to have it reproduced with gstreamer, but just wanted to ask if you have seen something like that before? [09:50:18 CEST] <fritsch> BtbN: ^^ [09:50:45 CEST] <fritsch> on the supplied images you can see that directly when comparing the green grass [09:50:51 CEST] <nevcairiel> you can run fate with vaapi if you want, that'll show you if its bitexact [09:50:54 CEST] <nevcairiel> which afaik it is [09:51:29 CEST] <nevcairiel> possibly your vaapi output chain isnt quite as good, if you use direct rendering to some degree [09:53:42 CEST] <fritsch> NV12 zero copy and then yvu2rgb shader [09:54:08 CEST] <fritsch> and disabling vaapi, using sw decoding and same yuv2rgb shader -> all details are there [09:54:38 CEST] <nevcairiel> just saying that you can test the decoder with copyback just by running "make fate-h264 HWACCEL=vaapi" [09:54:50 CEST] <nevcairiel> (some always fail, but that should be a minor subset, 1-3 or so) [09:55:06 CEST] <nevcairiel> if that passes, then something in how you use vaapi is to blame [09:55:15 CEST] <fritsch> okay, thx much - I will test that first [09:55:36 CEST] <nevcairiel> maybe its copy-back things are not quite as nice =p [09:55:41 CEST] <nevcairiel> eh, copy-less things [09:59:51 CEST] <jkqxz> The default settings there don't quite work. I use $(make HWACCEL='vaapi -vaapi_device /dev/dri/renderD128 -hwaccel_output_format yuv420p -hwaccel_lax_profile_check' fate-h264), which does. [10:00:14 CEST] <fritsch> what are the defaults? [10:00:14 CEST] <jkqxz> (Someone forgot to escape something properly, but it helps here...) [10:00:57 CEST] <fritsch> and what's the lax_profile_check? [10:01:25 CEST] <fritsch> and is yuv420p produced with ffmpeg or by vaapi - and how is the data fetched? [10:01:32 CEST] <nevcairiel> sounds like someone should fix the vaapi module to not suck [10:01:39 CEST] <fritsch> hehe [10:02:01 CEST] <jkqxz> It gives you the hardware format (probably NV12) and doesn't auto-convert for some reason I didn't pursue; it refuses to decode the all the baseline profile streams because they technically aren't supported. [10:02:07 CEST] <jkqxz> Yeah, probably :) [10:02:25 CEST] <fritsch> until some months ago vaapi had no way to get the NV12 data directly [10:02:26 CEST] <nevcairiel> giving you NV12 is fine, that still passes fate [10:02:34 CEST] <nevcairiel> dxva2 does that too [10:04:40 CEST] <ubitux> lol d1cacbbea [10:04:58 CEST] <jkqxz> s/convert /convert correctly /, perhaps? The yuv420p which comes out of vaapi is bit-exact, though, so I just use that and haven't investigated further. I guess I probably should now that other people can use it and will ask why its wrong... [10:08:11 CEST] <fritsch> Test h264-conformance-aud_mw_e failed. Look at tests/data/fate/h264-conformance-aud_mw_e.err for details. [10:08:17 CEST] <jkqxz> "and is yuv420p produced with ffmpeg or by vaapi - and how is the data fetched?" - on Intel the decoded surfaces are nv12, it gets downloaded by vaGetImage (i.e. VPP) to yuv420p, then maps the yuv420p buffer and copies to an AVFrame output. ffmpeg doesn't convert anything there. [10:11:46 CEST] <fritsch> jkqxz: okay - that we used too with sse4 intrinsics before we got this mesa / egl sharing [10:14:36 CEST] <fritsch> only fate-h264-conformance-cvfc1_sony_c seems to fail [10:14:43 CEST] <fritsch> jkqxz: same for you? [10:15:02 CEST] <wm4> nevcairiel: I didn't but that seems quite a specific issue [10:15:39 CEST] <wm4> err [10:15:46 CEST] <wm4> fritsch not nevcairiel [10:16:09 CEST] <fritsch> wm4: i let him test with mpv later on :-) [10:16:36 CEST] <fritsch> btw. howto skip a fate test to run the other tests? [10:16:46 CEST] <wm4> apt-get will usually install an ancient version btw. [10:16:58 CEST] <fritsch> if it works -> fine [10:17:01 CEST] <jkqxz> Sounds good. fate-h264-lossless always fails for me, and h264-conformance-frext-hpcvmolq_brcm_b / h264-conformance-frext-hpcamolq_brcm_b (the greyscale ones) do in some cases. [10:17:13 CEST] <nevcairiel> also the sony one [10:17:32 CEST] <fritsch> it seems for me it stops after the sony file [10:17:39 CEST] <jkqxz> Yeah, that one always fails. [10:17:41 CEST] <jkqxz> make -k ? [10:17:52 CEST] <fritsch> let me try - first time fade [10:17:52 CEST] <wm4> is the sony one cropping? [10:18:03 CEST] <jkqxz> Yes. Top/left cropping. [10:18:30 CEST] <fritsch> h264-lossless fails [10:18:31 CEST] <fritsch> yes [10:19:21 CEST] <fritsch> [vaapi @ 0x3f53e20] No VAAPI support for codec h264 profile 244: trying instead with profile 100. [vaapi @ 0x3f53e20] This may fail or give incorrect results, depending on your hardware. [10:19:27 CEST] <fritsch> yeah, seems that's planned to fail [10:19:44 CEST] <fritsch> as input is 4:4:4 [10:20:00 CEST] <nevcairiel> that seems unlikely, it would not use the hwaccel at all for 444 [10:20:38 CEST] <nevcairiel> lossless isnt necessarily 444 [10:20:47 CEST] <nevcairiel> that test in particular is 420 i think [10:21:14 CEST] <fritsch> Stream #0:0: Video: h264 (High 4:4:4 Predictive), yuv420p, 640x480, 25 fps, 60 tbr, 1200k tbn [10:21:52 CEST] <nevcairiel> thats just the profile name, the actual stream property is yuv420p [10:21:58 CEST] <nevcairiel> hence 4:2:0 [10:22:18 CEST] <nevcairiel> you can use any higher profile and only encode lower things in it [10:22:26 CEST] <nevcairiel> could use 10-bit profile and encode 8-bit content [10:22:30 CEST] <nevcairiel> just that noone does that [10:22:36 CEST] <nevcairiel> but there is no other lossless profile ;) [10:23:22 CEST] <BtbN> btw., we will be merging the libav and ffmpeg nvenc encoders, so both are the same. [10:25:16 CEST] <jkqxz> It's the -hwaccel_lax_profile_check option which is allowing it to even try there. Without it the hwaccel will reject the stream and use software - correctly, in this case. [11:34:29 CEST] <fritsch> okay, we do something wrong [11:34:34 CEST] <fritsch> in our sse copy path all fine [11:34:42 CEST] <fritsch> in our zero copy egl path -> issue [11:34:44 CEST] <fritsch> decoder is fine [11:34:54 CEST] <BtbN> My guess would be limited vs. full range rgb/yuv. [11:34:57 CEST] <fritsch> nope [11:35:03 CEST] <fritsch> that was already tested [11:35:11 CEST] <fritsch> same issue with limited / full [11:35:17 CEST] <fritsch> we get the data in limited [11:35:29 CEST] <BtbN> colorspace? [11:35:31 CEST] <fritsch> and depending on the output settings, it is scaled with dithering or not touched [11:35:35 CEST] <fritsch> both NV12 [11:35:45 CEST] <BtbN> That's not the colorspace though [11:35:52 CEST] <fritsch> and the yuv2rgb shader is the same for sse copy path and sw path [11:36:07 CEST] <fritsch> so the last render step is the same for all [11:36:10 CEST] <BtbN> Wouldn't be surprised if the EGL stuff in VAAPI messes with it [11:36:26 CEST] <fritsch> as it is "zero" copy i wonder [11:36:39 CEST] <BtbN> It's "don't pull it to system ram" [11:36:40 CEST] <fritsch> but that's what we will investigate next [11:36:56 CEST] <fritsch> i don't think it's the zero copy [11:37:06 CEST] <fritsch> as both times we use the same source [11:37:24 CEST] <fritsch> once copy with sse, otherwise use that mesa egl extension via the untouched buffer [11:37:43 CEST] <BtbN> do you have two sample images? One how it's supposed to look, and one how it ends up with EGL? [11:37:48 CEST] <fritsch> yes [11:38:11 CEST] <fritsch> http://pasteboard.co/PaVDs5i.bmp <- VAAPI with EGL [11:38:23 CEST] <fritsch> http://pasteboard.co/PbqgpjL.bmp <- some mediacodec whatever [11:38:32 CEST] <fritsch> see the face of the guy in the white shirt or the grass [11:38:56 CEST] <fritsch> vaapi with sse4 copy is fine - so it must be a step in the render which is in EGL but not in SW -> GL [11:49:41 CEST] <fritsch> that egl path looks correct [11:49:49 CEST] <fritsch> so it must be special part in LinuxRender [11:49:52 CEST] <fritsch> that happens only for this format [11:53:21 CEST] <wm4> good old bmp [11:54:30 CEST] <wm4> fritsch: could be something related to rendering? [11:54:36 CEST] <wm4> what exactly is the error description anyway [11:54:59 CEST] <fritsch> just look at the grass [11:55:04 CEST] <fritsch> error description enough :-) [11:55:09 CEST] <fritsch> or the face of the guy [11:56:03 CEST] <wm4> slightly more blurry, possibly different colors? [11:56:12 CEST] <fritsch> https://github.com/xbmc/xbmc/blob/master/xbmc/cores/VideoPlayer/DVDCodecs/V… wondering [11:56:46 CEST] <wm4> is vpp even involved at all? [11:56:51 CEST] <fritsch> no [11:56:56 CEST] <fritsch> not in that path [11:58:07 CEST] <fritsch> i wonder if we setup the surfaces in wrong format [11:58:13 CEST] <fritsch> and vaapi does an additional conversion [11:59:12 CEST] <fritsch> i linked the wrong part of the code [11:59:33 CEST] <fritsch> https://github.com/xbmc/xbmc/blob/master/xbmc/cores/VideoPlayer/DVDCodecs/V… <- that one is only happening [12:00:23 CEST] <wm4> yeah, that's quite straight-forward [12:00:32 CEST] <fritsch> yes [12:01:11 CEST] <fritsch> i need to check our LinuxRenderer code, cause that's the only position something is done before outputting [12:04:39 CEST] <fritsch> https://github.com/xbmc/xbmc/blob/master/xbmc/cores/VideoPlayer/VideoRender… [12:49:46 CEST] <fritsch> i think the dimensions do not match and the UV values degenerate when put on the surface [12:49:52 CEST] <fritsch> will prove it when at home [12:51:16 CEST] <wm4> cropping issue? [12:51:49 CEST] <wm4> I set EGL_WIDTH etc. of each plane to cropped values, and somewhat surprisingly that seems to work [14:34:30 CEST] <Daemon404> nevcairiel, i am sad such h264 streams exist [14:39:41 CEST] <Daemon404> how even runs fflogger btw? [14:39:43 CEST] <Daemon404> who* [14:42:18 CEST] <nevcairiel> Sounds like he did make clean, maybe some of the new objects are not being cleaned? [14:42:54 CEST] <Daemon404> read his latets comment [14:43:01 CEST] <Daemon404> he does not mention clean between builds [14:43:04 CEST] <Daemon404> only before the first of the two [14:43:26 CEST] <Daemon404> and object may not be rebuilt if it thinks it is targeting the same os and arch [14:43:29 CEST] <Daemon404> which he doesnt set [14:45:00 CEST] <wbs> Daemon404: it doesn't matter if configure thinks it's a different target; if the file itself doesn't include config.h (which not all object files do), they won't be rebuilt if configure updates config.h [14:45:23 CEST] <nevcairiel> He does mention make clean in the second step [14:51:43 CEST] <Daemon404> wbs, ah [14:51:49 CEST] <Daemon404> well then that explains it [14:52:24 CEST] <wbs> yeah, that's the telltale sign of forgetting to clean, if you get that error for some obscure table in lavc or obscure parts of lavu [14:52:44 CEST] <Daemon404> ive hit it enough that i now only build out of tree [14:52:47 CEST] <Daemon404> to save me pain [14:52:55 CEST] <Daemon404> in a dir per arch [14:53:09 CEST] <nevcairiel> Or clean lacking some entries because that's got refractories yesterday [14:53:17 CEST] <nevcairiel> tests* [14:53:29 CEST] <nevcairiel> Phone keyboard no fun [14:53:30 CEST] <Daemon404> i doubt that caused it [14:54:18 CEST] <Daemon404> hmm he make cleans after configure [14:54:23 CEST] <Daemon404> i guess that should still work [14:54:40 CEST] <Daemon404> i dont see how clean would have been affected by yesterdays tests tho [14:56:47 CEST] <Daemon404> daemon404@bbvm:~/dev/f/ffmpeg$ ls libavutil/*.o [14:56:47 CEST] <Daemon404> libavutil/adler32.o [14:56:47 CEST] <Daemon404> daemon404@bbvm:~/dev/f/ffmpeg$ make clean [14:56:47 CEST] <Daemon404> daemon404@bbvm:~/dev/f/ffmpeg$ ls libavutil/*.o [14:56:47 CEST] <Daemon404> ls: cannot access 'libavutil/*.o': No such file or directory [14:56:49 CEST] <Daemon404> daemon404@bbvm:~/dev/f/ffmpeg$ [14:56:52 CEST] <Daemon404> works fine here. [15:01:05 CEST] Action: nevcairiel tests the exact commands [15:01:17 CEST] <Daemon404> i dont have msvc installed atm [15:13:10 CEST] <Daemon404> :) [15:16:38 CEST] <cone-646> ffmpeg 03Anton Khirnov 07master:06edef3d5e07: Generate the lists of enabled protocols/bsfs from configure. [15:16:39 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:9f8a942d6ac9: Merge commit '06edef3d5e072ef3c4face9ce946d2d9c36cc477' [15:18:16 CEST] <cone-646> ffmpeg 03Thomas Guillem 07master:785bfb1d7bb8: pixfmt: fix wrong comment [15:18:17 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:eae2ebded3b8: libxvid: Create extradata in init using a dummy frame [15:18:18 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:92d934106cb2: Merge commit '785bfb1d7bb8de567c3aac1d9cc369b55ac9fb7b' [15:18:19 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:f8a4b8bc4059: Merge commit 'eae2ebded3b801ed55d32746b98db88ffe196f4f' [15:21:24 CEST] <cone-646> ffmpeg 03Luca Barbato 07master:6b2ad3ca48a6: indeo3: Avoid undefined behaviour [15:21:25 CEST] <cone-646> ffmpeg 03Luca Barbato 07master:f3fdef108eb0: ape: Avoid undefined behaviour [15:21:26 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:3fd5f09845e5: Merge commit '6b2ad3ca48a6638cb0226ed5aab41d435d8c83a5' [15:21:27 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:22900770c56d: Merge commit 'f3fdef108eb06b1e71b29152bf6822519e787efe' [15:28:38 CEST] <cone-646> ffmpeg 03Luca Barbato 07master:79fdbfdb3e50: img2enc: Refactor the atomic renaming code [15:28:39 CEST] <cone-646> ffmpeg 03Martin Storsjö 07master:0abb07bad702: movenc: Update a comment to reflect how the code actually behaves [15:28:40 CEST] <cone-646> ffmpeg 03Martin Storsjö 07master:74383def8f46: movenc: Handle pts == NOPTS when autoflushing [15:28:41 CEST] <cone-646> ffmpeg 03Martin Storsjö 07master:75b90ef722b7: libavformat: Update the comment about AVOutputFormat flags [15:28:42 CEST] <cone-646> ffmpeg 03Diego Biurrun 07master:a08b5d7b5725: build: Silence the lcov-reset target [15:28:43 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:a2906846da10: Merge commit '79fdbfdb3e50f3f906903e027714ee04c1a00e89' [15:28:44 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:a022c1fe76f3: Merge commit '0abb07bad7026a945a31ba4047e6583c8b3fa3da' [15:28:45 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:ee96b7b1c656: Merge commit '74383def8f46805faf3391c98516b248108a9a6b' [15:28:47 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:1549ca42672f: Merge commit '75b90ef722b7cdfc70118ab987e298d087aae693' [15:28:47 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:d4acf9f2dd2a: Merge commit 'a08b5d7b5725932f4ad39e95c5d6589392dee2c6' [15:29:45 CEST] <cone-646> ffmpeg 03Vittorio Giovara 07master:9e2af0e9071a: libx264: Allow Stereo3D monoscopic value [15:29:45 CEST] <cone-646> ffmpeg 03Vittorio Giovara 07master:5fca95c8e515: libx264: Forbid inverted Stereo3D mode [15:29:46 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:e3cfd1b27515: Merge commit '9e2af0e9071a1229cfe21efff394691d91f979b2' [15:29:48 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:5b9a4476e312: Merge commit '5fca95c8e515a5ae542d9626ec088bdfc658450e' [15:30:13 CEST] <Daemon404> just reached the Giant H264 Set [15:30:14 CEST] <Daemon404> :V [15:31:01 CEST] <Daemon404> oh goddammit [15:31:04 CEST] <Daemon404> https://git.libav.org/?p=libav.git;a=commit;h=4012fe1ee819edc7689e182189e66… [15:32:06 CEST] <nevcairiel> just skip that shit, its one of those cases were simply porting the solution we had was not good enough, even though it was directly shoved into their faces =p [15:32:25 CEST] <nevcairiel> they even had to do the same mistake :) [15:32:39 CEST] <cone-646> ffmpeg 03Luca Barbato 07master:dd4fb2339f76: ape: Unbreak adaptcoeffs computation [15:33:01 CEST] <Daemon404> nevcairiel, too late now [15:33:03 CEST] <Daemon404> so i picked the fix [15:36:50 CEST] <cone-646> ffmpeg 03Anton Khirnov 07master:f3ed484953b8: h264_mp4toannexb_bsf: do not fail on annex B extradata [15:36:52 CEST] <cone-646> ffmpeg 03Derek Buitenhuis 07master:97946b20b6d8: Merge commit 'f3ed484953b81856e40239d2410058a96188b2be' [15:37:15 CEST] <Daemon404> nevcairiel, should i hold off on the rest until you push your other patches to h2645 [15:38:07 CEST] <nevcairiel> dont think those touch h2645 [15:38:24 CEST] <Daemon404> sure but [15:38:34 CEST] <Daemon404> h264_parser: switch to h2645_parse for NAL unescaping [15:39:32 CEST] <nevcairiel> all i know is that i need to update a bunch of patches in my lav fork aftrer all this shit [15:40:31 CEST] <Daemon404> i dont think youll find much sympathy for that =p [15:41:17 CEST] <BtbN> At least future nvenc merges should be trivial, or rather non-existent. [15:42:12 CEST] <nevcairiel> you say that now, but syncing up once doesnt guarantee staying in sync for ever :D [15:42:48 CEST] <BtbN> That's what we're planning though, keeping them in sync. [15:43:18 CEST] <BtbN> We are currently merging the ffmpeg and libav one, using the libav encoder as base, because we found it to look nicer. [15:44:06 CEST] <wm4> lol [15:45:14 CEST] <nevcairiel> "look" nicer, but ours has like 100 more features [15:45:27 CEST] <nevcairiel> looks are more important then, i see [15:45:49 CEST] <BtbN> I won't accept the merge if the result is a step back for anything. [15:45:57 CEST] <wm4> you seem quite disenchanted with Libav recently [15:46:21 CEST] <BtbN> So no lost features, no changes in how it's invoked. [15:47:32 CEST] <BtbN> There are numerous things on both sides that would have been needed to merge to the other. [15:57:31 CEST] <Daemon404> nevcairiel, urg [15:57:42 CEST] <Daemon404> merging that commit seems to break every single h264 file in fate [15:57:46 CEST] <Daemon404> i must have screwed up... [15:58:00 CEST] <Daemon404> oh not every single one [15:58:02 CEST] <Daemon404> just a bunch [15:58:03 CEST] <Daemon404> but still. [15:58:53 CEST] <nevcairiel> did I mention that i'll be mostly out this weekend [15:59:16 CEST] <Daemon404> lol. [15:59:27 CEST] <Daemon404> thats fine. no merges till next week then [15:59:28 CEST] Action: Daemon404 runs [15:59:48 CEST] <nevcairiel> the commit looks rather simple though [16:00:27 CEST] <Daemon404> maybe our code had some special case? [16:05:33 CEST] <nevcairiel> ours supports mp4-style, theirs doesnt, so might need some adjustments [16:08:02 CEST] <Daemon404> why would that affect the conformance tests in fate [16:08:06 CEST] <Daemon404> arent they all annexb [16:08:13 CEST] <nevcairiel> dunno [16:08:16 CEST] <nevcairiel> maybe you broke some more [16:08:17 CEST] <nevcairiel> :) [16:08:35 CEST] Action: Daemon404 pawns off responsibility to nevcairiel next week and does work-work instead [16:09:10 CEST] <nevcairiel> i can maybe look at it sunday [16:09:28 CEST] <nevcairiel> but today is thursday, so that means cinema, and tomorrow i'm already leaving early [16:10:11 CEST] <Daemon404> enjoy [16:10:48 CEST] <Daemon404> and now unrelated [16:11:01 CEST] Action: Daemon404 should really unfollow all these EBU and broadcaster people on twitter [16:11:21 CEST] <Daemon404> all they ever post is stuff from circlejerk conferences that pat their own backs [16:12:17 CEST] <nevcairiel> well then do so [16:12:20 CEST] <nevcairiel> its not like thats a lot of work [16:12:46 CEST] <Daemon404> you underestimate my laziness [16:15:13 CEST] <nevcairiel> if its anything like my own... [16:19:12 CEST] <Daemon404> nevcairiel, fyi applying your patch doesnt fix anything [16:20:16 CEST] <Daemon404> http://chromashift.org/merge.txt [16:20:17 CEST] <Daemon404> ^ how i merged [18:26:28 CEST] <pfelt> afternoon all. i've got a question on commits again. so i submitted a patch and didn't get anythign back on the -devel list. turns out it was committed the next day and i just didn't know it. apologies for bumping it on the list. the question is, what's the right way to be notified of it? only way i can think of is to watch ffmpeg-commits (which i'll do if that's the right way) [18:28:43 CEST] <wm4> pfelt: normally whoever applied it will reply to your mail (on the list) when it's pushed [18:28:49 CEST] <wm4> maybe it was just forgotten this time [18:30:59 CEST] <jkqxz> It looks like michaelni applied your patch but replied to <http://ffmpeg.org/pipermail/ffmpeg-devel/2016-May/194126.html> instead? [18:32:32 CEST] <jkqxz> Unclear which way around that was confused, though. (Was it actually meant to be the other patch which was applied, or was the response in the wrong thread?) [18:37:27 CEST] <pfelt> wm4: ok :) totally not a problem. just wanted to make sure i was doing things the right way [18:38:48 CEST] <wm4> you could of course also watch the git master branch [18:46:49 CEST] <pfelt> ok. another noob question. i just local committed and generated a new patch for submission and i found i left DISABLE_DEPRECATION_WARNINGS in two places where those shold not be committed. do i need to unstage and restage or can i just remove those two hunks from the patch? [18:50:43 CEST] <wm4> the easiest way I found to "edit" commits in a fine grained unannoying way so far is using git-cola [18:50:58 CEST] <wm4> and then just click amend and unstage the unwanted lines [18:53:47 CEST] <pfelt> :( ok [18:53:49 CEST] <BtbN> undo those changes, then commit --amend, and then checkout the file from the overwritten commit(note the ID before overwriting it) [18:54:25 CEST] <pfelt> i saved the patch out before i committed so i just branched a new local branch, re-applied the patch and i'm restaging it int he new branch [18:55:06 CEST] <wm4> using git-cola has the advantage that you don't actually need to remove the changes or reset the commit [18:56:54 CEST] <pfelt> i only cursorily looked at it, but it's a gui so i assumed i needed x installed. i don't currently [18:58:19 CEST] <jkqxz> I use rebase -i / commit --amend. Works nicely when the edits are small enough that you can make them live, but annoying once you want to make larger changes to a commit series (use stash in that case). [19:00:57 CEST] <Shiz> ^same [19:02:20 CEST] <pfelt> in this case it was removing two complete hunks (4 added lines in each) both in the same file [20:54:36 CEST] <Daemon404> this guys results arent making any sense [21:56:04 CEST] <BBB> I think his results make perfect sense [21:57:45 CEST] <Daemon404> BBB, i was thrown off my adler32.o [21:57:54 CEST] <Daemon404> i see it now with his latest comment [21:58:15 CEST] <BBB> I think it complains about the first non-compat file, which happens to be adler32.o [21:58:38 CEST] <BBB> it probably links in this order: compat/*.o (all bad arch) [rest] with adler32 being first in [res] [21:58:44 CEST] <Daemon404> yeah exactly [21:58:45 CEST] <BBB> and then complains that adler32 is broken [21:58:50 CEST] <BBB> because its not the same as compat/*.o [21:58:54 CEST] <BBB> anyway [21:59:01 CEST] <BBB> easy for me to talk after he figured out the issue :-p [21:59:16 CEST] <Daemon404> i only figured it out when he listed the files in the latest one [21:59:23 CEST] <BBB> I should invent a time travel machine and show you guys how clever I am ten minutes ago [21:59:28 CEST] <Daemon404> lul [22:00:19 CEST] <BBB> hey kierank, are you asleep? [22:01:16 CEST] <Daemon404> at 9pm? [22:01:32 CEST] <BBB> you never know, he might be a leicester city fan [22:01:55 CEST] <kierank> In the pub actually [22:02:07 CEST] <Daemon404> are you ever not in a pub after 7pm [22:02:40 CEST] <BBB> didnt he mention he was working an office job, effectively? [22:02:46 CEST] <BBB> so pub@7 is actually very reasonable [22:02:47 CEST] <Daemon404> mind you i'd probably get a drink at the local after work if my local wasnt a 20 minute walk away [22:02:51 CEST] <BBB> better than pub@4:30 [22:03:08 CEST] <Daemon404> BBB, i have a lunch beer instead of a beer after work actually [22:05:01 CEST] <kierank> BBB: yes? [22:05:22 CEST] <BBB> do you mind a privmsg chat? or are you typing on your phone? :-p [22:05:30 CEST] <kierank> On phone [22:06:14 CEST] <kierank> I'll stop being antisocial [22:06:50 CEST] <BBB> haha, well talk later then [22:08:10 CEST] <Daemon404> actually im a hypocrite [22:08:18 CEST] Action: Daemon404 didnt even realize he was drinking beer at this moment [22:08:44 CEST] <BBB> I should get a beer [22:10:48 CEST] <BBB> but I think Im having sake pairing tonight [22:10:51 CEST] <BBB> so Ill go with that [22:10:53 CEST] <BBB> yummy [22:12:25 CEST] <Daemon404> http://www.beerritz.co.uk/buy/wild-beer-co-the-blend-summer-2015-75cl_2260.… <-- my drink [22:23:09 CEST] <fritsch> "wild beer" [22:25:15 CEST] <wm4> fritsch: so what was that vaapi problem? [22:25:56 CEST] <fritsch> cropping [22:26:03 CEST] <fritsch> but need to fix it first :-) [22:26:07 CEST] <Shiz> fritsch: better name than america [22:26:33 CEST] <fritsch> wm4: only happens with 1920x1088 files [22:26:52 CEST] <fritsch> we mix some Y - UV correspondences [22:27:16 CEST] <fritsch> but evenen is nearly over, cannot even remember how I wasted that time [22:27:39 CEST] <fritsch> Shiz: yeah - seen that ... not sure what they thought about doing that name change [22:27:56 CEST] <fritsch> but when reading "wild beer" I always think if people know how one makes beer ... [22:28:14 CEST] <fritsch> and there is nothing "wild" left ... when it's brewed [22:28:24 CEST] <Shiz> artisan hand-plucked beer [22:28:29 CEST] <Shiz> from the finest beer trees [22:28:47 CEST] <fritsch> that's how it sounds :-) [22:29:43 CEST] <Daemon404> fritsch, their name comes from the "wild ale" style of beer [22:30:10 CEST] <Daemon404> i think. [22:30:39 CEST] <fritsch> https://en.wikipedia.org/wiki/Reinheitsgebot [22:30:48 CEST] <fritsch> ^^ my comment to such jokes [22:30:49 CEST] <fritsch> :p [22:30:56 CEST] <Daemon404> yes i know [22:31:04 CEST] <Daemon404> i make fun of that law all the time [22:31:13 CEST] <Daemon404> theres so many offensive jokes to make. [22:31:36 CEST] <fritsch> i just found out there is a "Simple English" [22:31:39 CEST] <fritsch> wikipedia language [22:32:24 CEST] <fritsch> how nice of them ... https://simple.wikipedia.org/wiki/Simple_English_Wikipedia [22:32:34 CEST] <fritsch> small sentences, simpler words and grammar [22:32:47 CEST] <fritsch> let's see if they translated some math stuff to that language [22:35:44 CEST] <Daemon404> https://simple.wikipedia.org/wiki/Fourier_transform [22:35:53 CEST] <Daemon404> i dont think it is a accompishing its goal [22:36:46 CEST] <fritsch> hehehe, in deed [22:55:04 CEST] <Shiz> lol [23:59:41 CEST] <pfelt> if anyone is on reviewing my updated patch to decklink, hold off. i just learned a valuable lesson in that i see another issue that i *should* have caught before submitting another rev [00:00:00 CEST] --- Fri May 13 2016

1 0

ffmpeg.log.20160512
by burek 13 May '16

13 May '16

[00:34:00 CEST] <pfelt> are there currently any output formats that run multithreaded? [00:34:18 CEST] Action: pfelt is playing with the decklink output format and it appears to be single threaded [00:34:43 CEST] <JEEB> muxers aren't usually threaded [00:34:48 CEST] <JEEB> encoders are [00:35:42 CEST] <pfelt> hmm. so if i have the following command line it's going to be single threaded regardless as to what format i use? [00:36:32 CEST] <pfelt> -map '[ulf]' -f decklink 'DeckLink Quad (3)' -map '[urf]' -f decklink 'DeckLink Quad (7)' [00:36:58 CEST] <JEEB> no idea about decklink stuff or what encoder gets used with that if any [00:37:16 CEST] <JEEB> post your terminal output [00:37:22 CEST] <pfelt> it's output is from rawvideo [00:37:24 CEST] <JEEB> in pastebin-like :P [00:37:28 CEST] <pfelt> heh. yeah [00:37:37 CEST] <c_14> That should probably generate at least 2 threads. One per output afaik [00:38:27 CEST] <JEEB> yes, I would expect a single muxer be a single thread there, while decoders and encoders and filters have their own limitations and capabilities [00:39:04 CEST] <pfelt> http://pastebin.com/PzZeNkbm [00:39:06 CEST] <pfelt> that enough ? [00:39:41 CEST] <pfelt> so i'm willing to sweat out making decklink mutlithread if there is already an example somewhere [00:40:09 CEST] <pfelt> and i found that it's single thread because i can't really run more than one output without dying unless i use named pipes and lots of ffmpeg processes (one per output) [00:40:25 CEST] <pfelt> gdb break on the output packet write seems to always be in the same thread [00:40:33 CEST] <JEEB> multithreading sounds really weird in a muxer [00:40:47 CEST] <JEEB> since usually a muxer isn't something you thread, it's just IO work [00:41:21 CEST] <pfelt> really i just need a way to output to multiple outputs at near realtime and synchronized pts [00:41:24 CEST] <JEEB> having it in a separate thread and constantly feeding it stuff... sure. also you have two /dev/null outputs there but otherwise looks OK [00:41:31 CEST] <pfelt> it's quite difficult to keep the named pipes in sync [00:41:52 CEST] <pfelt> yeah. if i change those nulls to two more outputs it's unreasonably slow [00:42:19 CEST] <pfelt> (even with just the two it eventually slows down so much where the decklink card is waiting for frames because ffmpeg can't get them there fast enough) [00:42:30 CEST] <JEEB> also you might want to try a filter chain where you convert to the output format before splitting [00:42:32 CEST] <pfelt> pts gets all hosed up and the video slows down to something like 1/3 actual speed [00:42:35 CEST] <JEEB> unless you're already doing it [00:43:18 CEST] <JEEB> otherwise I'm not sure if threading is the thing to do in the muxer, but you might want to look into optimizing the handling of the decklink stuff, I guess? [00:43:20 CEST] <pfelt> idea is to cut the video stream into 4 quarters and put one quarter out each output [00:43:47 CEST] <pfelt> any ideas on how to best find where the bottleneck is in ffmpeg? [00:44:30 CEST] <JEEB> start with decoding, then add one part of your filter chain (the common one), then add the cutting and then add the outputs [00:45:19 CEST] <JEEB> also as I noted, you might want to do it so that you do source->pix_fmt conversion to the required format (seems to be uyvy422?)->cutting [00:45:36 CEST] <JEEB> that way all those outputs don't separately convert their parts into that pix_fmt [00:47:39 CEST] <pfelt> i had thought about doing that, but i think crop requires rgp or at the least a different uyv format [00:47:43 CEST] <pfelt> i can retry it [04:21:37 CEST] <prelude2004c_Zzz> hey guys... i am having some trouble.. i was suggested to use fifo to do the job of a playout list and then a listener to segment.. ( see code here : http://pastebin.com/raw/aMF0uM5j ) .. but its acting up. sometimes the sources just stop going and i am having trouble with stability [04:21:39 CEST] <prelude2004c_Zzz> can anyone help ? [04:21:43 CEST] <prelude2004c_Zzz> is there something here i did wrong ? [04:46:01 CEST] <prelude2004c_Zzz> anyone? [04:47:33 CEST] <mundus2018> ? [05:39:20 CEST] <thebombzen> question about hardsubbing - so I want to hardsub the video but only after the first 10 minutes [05:39:35 CEST] <thebombzen> so I use ffmpeg -ss -i input.mkv -vf subtitles=input.mkv [05:40:04 CEST] <thebombzen> sorry I mean I use ffmpeg -ss 10:00 -i input.mkv -vf subtitles=input.mkv options out.mkv [05:40:19 CEST] <thebombzen> but that causes the subtitles to be misaligned [05:40:46 CEST] <thebombzen> i.e. the first 10 minutes of the subtitles overlay on where the video starts. how do I start the subtitles filter 10 minutes into the video [05:41:18 CEST] <rsully> not sure, but depending on subtitle format you may be able to edit them so that there aren't any subs until the 10 min mark [05:41:59 CEST] <rsully> but I generally recommend not hardsubbing - add as a soft track and set forced flag or not [05:42:09 CEST] <thebombzen> the goal is to upload to YouTube [05:42:12 CEST] <thebombzen> which doesn't support softsubs [05:42:19 CEST] <rsully> youtube supports captions [05:42:25 CEST] <thebombzen> I mean I could always do ffmpeg -ss 10:00 -i input.mkv -map 0:s -c ass out-cropped.ass but that's not what I'm looking for [05:42:27 CEST] <rsully> you can upload separate from video [05:42:31 CEST] <thebombzen> also youtube closed captioning is terrible [05:42:39 CEST] <thebombzen> I have pretty ASS subs that I want to be pretty [05:42:41 CEST] <rsully> not when it is supplied by you [05:42:51 CEST] <rsully> oh i thought you meant their speech to text [05:43:13 CEST] <rsully> well, as a viewer I hate hard subs [05:43:36 CEST] <rsully> I'd rather have ugly than encoded into the video stream [05:43:51 CEST] <rsully> not that helvetica is ugly [05:43:53 CEST] <thebombzen> so in general. if someone asks how to do something. and you don't know know why they're asking [05:44:01 CEST] <thebombzen> your answer should not be "don't do it" [05:44:27 CEST] <thebombzen> if you don't have any actual advice to give other than "don't do it" then please don't tell me not to [05:44:35 CEST] <rsully> this is community support man. you have any idea how many people here ask things and they don't realize there are other ways to do it? you have to prove your knowledge if you don't want my canned responses [05:45:05 CEST] <thebombzen> idk it just pushes my buttons when I ask questions on the internet and people are like "why would you want to do that?" well it doesn't actually matter why. clearly I'm asking for a reason. [05:45:06 CEST] <rsully> if anyone else knew they'd chime in [05:45:28 CEST] <thebombzen> suppose I wanted to give it to a friend who uses iMovie (which doesn't support ASS afaik). [05:45:39 CEST] <thebombzen> suppose I want it to run on my grandmother's computer and she uses Windows Media Player [05:45:48 CEST] <rsully> most people don't have good reasons, they just have a path they think they need to follow [05:45:58 CEST] <rsully> does WMP even support mkv? [05:46:07 CEST] <thebombzen> probably not. but that's not the point. [05:46:29 CEST] <rsully> hm I wonder if final cut supports ASS [05:46:40 CEST] <rsully> I'd imagine imovie would support the same [05:46:56 CEST] <thebombzen> that's also not the point [05:47:27 CEST] <thebombzen> the point is if someone asks how to do something on the internet and you don't know why they want to do it, then don't say "don't do that". just keep silent [05:47:35 CEST] <thebombzen> because it's really annoying [05:47:37 CEST] <rsully> all my point is is that I didn't know your motives, but I know a lot of people ask for what you want when it isn't actually what they want [05:48:02 CEST] <thebombzen> another pet peeve - answering the question they think you mean to ask rather than the question you asked. [05:48:06 CEST] <rsully> too bad youtube doesn't support PGP or ASS [05:48:08 CEST] <thebombzen> don't do that. [05:48:18 CEST] <furq> i will carry on doing both of those things [05:48:21 CEST] <rsully> again, community support. tough luck. [05:48:36 CEST] <furq> more often than not people don't actually know what they want [05:49:03 CEST] <furq> anyway i would expect the easiest way would just be to remove the first ten minutes from the ass [05:49:05 CEST] <thebombzen> anyway, the original question still stands. is there some kind of hidden option in the subtitles filter that simulates -ss or do I have to manually remux. [05:50:22 CEST] <furq> you could try -vf subtitles=foo.ass:enable=gt(t\,600) [05:50:29 CEST] <furq> i don't know if the subtitles filter supports timeline editing though [05:51:01 CEST] <furq> if that doesn't work then you'll have to do it manually [10:26:51 CEST] <ferdna> i need to open 16 streams from the same feed.. [10:27:01 CEST] <ferdna> if i do that it freezes [13:01:41 CEST] <theeboat> Does anybody know if it is possible to extract the timecode from a h264 stream being sent over udp. I am told that the timecode is stored in the SEI pic_struct and I would like to view it in the following format HH:MM:SS:FF. thanks [13:02:14 CEST] <c0rnw19> Bonjour les gars! [13:32:56 CEST] <hanDerPeder> is it possible to 'pause' a stream opened with avformat_open_input? or should I close the stream and reopen later? reading from an internet radio url [14:07:16 CEST] <theeboat> Does anybody know if it is possible to extract the timecode from a h264 stream being sent over udp. I am told that the timecode is stored in the SEI pic_struct and I would like to view it in the following format HH:MM:SS:FF. thanks [14:08:52 CEST] <DHE> usually UDP h264 is in mpegts format (UDP packet payload size of 1316 bytes) [14:11:57 CEST] <jkqxz> The full time information in the SEI is just discarded; see <http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavcodec/h264_sei.c;hb=HEA…>. [14:41:17 CEST] <Aerroon> hello, i have a problem with what i presume to be a recording issue but i was hoping that someone here could potentially push me towards the right direction to seek an answer (i'm recording with obs using quicksync) [14:41:49 CEST] <Aerroon> sometimes the recordings have the start of the recording be artifacted heavily, when i play it back in VLC the first few seconds look like this: https://dl.dropboxusercontent.com/u/89115692/wows4/weirdness.png [14:42:26 CEST] <Aerroon> if i demux said recording with ffmpeg those first seconds are completely cut from the video [14:42:40 CEST] <Aerroon> only error thrown is this: [h264 @ 0000000002293b00] co located POCs unavailable [14:43:01 CEST] <Aerroon> my mpc-hc setup playing it back just shows the first frame for those seconds and then plays on normally: what could this be? [14:46:01 CEST] <Aerroon> i know that this isn't exactly an ffmpeg issue but perhaps someone here knows why it could be happening or if i could somehow make ffmpeg preserve the first few seconds when demuxing in however messed up way they stay in so it'd be easier to sync stuff [14:49:54 CEST] <Aerroon> also i'll be around so if you have an answer just highlight me even if it's hours later [14:51:01 CEST] <kbo> Hi, I have a weird issue after cross-compiling ffmpeg, when I try to execute ffmpeg binaries I have a "No such file or directory", I've already checked the binary format which correspond to what I expected (ARM EABI), any help is welcome ! [14:52:19 CEST] <kbo> here the configure command: ./configure --cross-prefix=arm-poky-linux-gnueabi- --enable-cross-compile --target-os=linux --arch=arm --cpu=cortex-a8 --sysroot=/home/kbo/sdk/sysroots/cortexa9hf-vfp-neon-poky-linux-gnueabi --enable-libx264 --enable-gpl [14:53:02 CEST] <kbo> --extra-cflags=' -march=armv7-a -mfpu=neon -mfloat-abi=hard' [14:54:48 CEST] <jkqxz> kbo: That's coming out immediately with no other output? Probably missing something the dynamic linker wants to find on the target - try running it with LD_DEBUG set to see what it's not finding. ('LD_DEBUG=all ./ffmpeg') [14:56:33 CEST] <kbo> jkqxz: I got the same result : root@imx6:~# LD_DEBUG=all ffmpeg -sh: /usr/bin/ffmpeg: No such file or directory root@imx6:~# ldd /usr/bin/ffmpeg /usr/bin/ldd: line 116: /usr/bin/ffmpeg: No such file or directory [14:57:57 CEST] <kbo> jkqxz: it's like ffmpeg isn't linked against any libs [15:03:05 CEST] <jkqxz> kbo: Do you even have the dynamic linker it asks for in executable? [15:04:25 CEST] <kbo> jkqxz: how can I know if i have it ? [15:05:34 CEST] <jkqxz> Look at 'readelf -l ./ffmpeg'. It will mention a program interpreter '/libsomething/ldsomething'. Does that actually exist on your target? [15:07:11 CEST] <kbo> result : [Requesting program interpreter: /lib/ld-linux.so.3] [15:07:46 CEST] <jkqxz> And can you run '/lib/ld-linux.so.3' on the target? [15:09:11 CEST] <kbo> jkqxz: ok, it actually work with the command /lib/ld-linux-armhf.so.3 /usr/bin/ffmpeg [15:09:41 CEST] <kbo> jkqxz: I got: ffmpeg version 2.2.3 Copyright (c) 2000-2014 the FFmpeg developers built on Mar 23 2016 13:48:27 with gcc 4.9.1 (GCC) ... [15:10:47 CEST] <jkqxz> Sounds like you have a soft-float toolchain and a hard-float target. It might work a bit, but will probably barf horribly (or just give wrong results for everything) if you ever try to use floating point. [15:14:18 CEST] <jkqxz> So, use a toolchain which actually matches the target machine. [15:14:58 CEST] <kbo> jkqxz: ok so its related to toolchain configuration, thanks for you help [17:05:30 CEST] <ferdna> good morning [17:05:48 CEST] <ferdna> i need help with ffserver [17:06:00 CEST] <ferdna> i amn trying to stream one source to many clients [17:06:39 CEST] <ferdna> the problem is that i run out of bandwidth.... [17:06:45 CEST] <ferdna> or something [17:06:47 CEST] <ferdna> not really sure [17:18:15 CEST] <EightBitSloth> I'm running a local nginx-rtmp server and I'd like to take it's output, which is at rtmp://localhost/live/test, and redirect it to v4l2loopback at /dev/video1. My goal is to be able to stream my desktop as a webcam when a user does not have access to my local nginx server. Can ffmpeg do this? [17:19:51 CEST] <thebombzen> you can always stream your desktop with ffmpeg's x11grab input device [17:20:14 CEST] <DHE> yes, but it sounds like a choice between "use my own RTMP server" and "stream with some kind of webcam software" [17:20:18 CEST] <DHE> trying to cheat a bit and do both [17:20:31 CEST] <thebombzen> ffmpeg -f x11grab -framerate 60 -video_size 1920x1080 -i :0.0 is an example [17:20:34 CEST] <EightBitSloth> Yep, but I need to be able to transition between desktop and applications. [17:21:03 CEST] <thebombzen> EightBitSloth: is that what you're looking for? [17:22:19 CEST] <EightBitSloth> @thebombzen, Not exactly. I need to do transitions and that doesn't allow me to. I forgot to add I need to use OBS. [17:22:20 CEST] <theeboat> jkqxz: thanks for that link, that makes more sense to me now. Do you know of any other software which writes the SEI time information or if it is possible to do this is ffmpeg? It seems a bit strange that this information would be dropped as it is quite a useful bit of information. [17:22:53 CEST] <EightBitSloth> I've streamed video to loopback and my desktop to loopback, so I figured sending an rtmp stream to loopback would work too. [17:27:59 CEST] <EightBitSloth> Oh hey! Found it myself. Apparently I just need to add rawvideo and change localhost to my local ip. [17:28:16 CEST] <EightBitSloth> ffmpeg -i "rtmp://streamurl" -vcodec rawvideo -y -f v4l2 /dev/video1 [17:34:19 CEST] <jkqxz> theeboat: Have you got a stream which actually contains it? The encoder certainly isn't obliged to supply it at all, and I don't know of any which do. (x264 never writes it, for example.) [17:37:16 CEST] <jkqxz> (Here is x264 never writing it in pic_timing SEI: <http://git.videolan.org/?p=x264.git;a=blob;f=encoder/set.c;hb=HEAD#l618>.) [17:38:47 CEST] <theeboat> jkqxz: I am using a hardware encoder, i have spoken to an engineer for the company of the product and he said the following. The time code goes into the SEI pic_struct as defined in the H.264 specification. [17:38:57 CEST] <vade> where does AudioFrameQueue store the actual samples it enqueues from the AVFrame you add? Im trying to understand how to use the ff_af_queue_* API to ensure my encoder gets AVFrames of the right frame size [17:39:20 CEST] <vade> but I cant understand how to pull frames off of the AudioFrameQueue to make a new AVFrame* with the right number of samples [17:39:46 CEST] <ddmd> Anyone know how I can change a video's color space from YUV to RGB? [17:40:20 CEST] <kepstin> ddmd: libswscale? (if you want a better answer, you need more detail in your question) [17:40:31 CEST] <vade> ddmd: have to use a SwsContext via libswscale [17:41:05 CEST] <kbarry> http://stackoverflow.com/questions/27519056/warning-in-converting-yuv-color… [17:41:38 CEST] <kbarry> Looks like that might offer some help with your problem ddmd [17:42:40 CEST] <jkqxz> theeboat: You probably just want to write the code in ffmpeg to store it somewhere you can read, then. Adding it (replacing the code in the previous link) would not be difficult, I think. [17:44:47 CEST] <ddmd> Thanks guys. I've got an mp4 playing in a html video tag and internet explorer displays white as gray. I've read that it could be the colorspace and that I should use BT.709. Does that make any sense? [17:44:49 CEST] <theeboat> jkqxz: OK, thanks for the information. I was hoping to not have to change any code but if that is the only option. The comment in the x264 commit makes more sense into why it would be dropped. [17:45:35 CEST] <kepstin> ddmd: you don't want to use rgb, most web browsers can only decode yuv. The colorspace is something separate from that. [17:46:48 CEST] <kepstin> ddmd: are you encoding a video using the ffmpeg command-line tool? [17:47:08 CEST] <ddmd> So if not the color space, what could throw the white balance off like that? Yes, I'm using ffmpeg for everything [17:47:42 CEST] <kepstin> it might be the colorspace. but you're getting that confused with the pixel format, something rather different [17:48:42 CEST] <kepstin> if you're getting white showing up as grey, it could actually be the sample range rather than color space. What's the original video source? [17:49:39 CEST] <ddmd> Its originally an mp4. I don't have unprocessed video. [17:50:42 CEST] <vade> post the ffprobe output perhaps? [17:50:56 CEST] <kepstin> anyways, you can use the "scale" video filter in ffmpeg to convert between colorspaces and sample ranges. Try "-vf scale=in_color_matrix=bt601,out_color_matrix=bt709" to convert the video color space, maybe? [17:52:20 CEST] <ddmd> here is some more info: http://pastebin.com/vrTLLFkP [17:52:58 CEST] <kepstin> that said, the white point should be the same for bt601 and bt709, iirc? [17:53:19 CEST] <ddmd> Thanks. I'll try out scale/ [17:55:45 CEST] <vade> how to deal with more samples than frame size ? I get that my AVFrame that ive converted with libswresample has more samples in it than the encoder is wanting. Whats the ffmpeg API correct way to handle that? [18:06:09 CEST] <ddmd> I get "[AVFilterGraph @ 0473e560] No such filter: 'out_color_matrix' " is this not installed by default? [18:06:09 CEST] <vade> im also flagging AV_CODEC_CAP_VARIABLE_FRAME_SIZE, but my audioStreams codec context lists 1024 [18:06:27 CEST] <vade> even though I set it to 0 and mark that flag prior to opening the codec [18:26:33 CEST] <vade> :\ [18:31:08 CEST] <durandal_1707> ddmd: colorspace filter [18:31:51 CEST] <spoon> Hi, I am trying to decode an RTP stream with the C API but getting pretty much nowhere [18:32:36 CEST] <spoon> My main issue is I have my own RTSP layer, and this makes things difficult merge with FFMPEG [18:34:30 CEST] <spoon> I require a lot of help unfortunately [18:37:58 CEST] <marsupial> can someone help me? i've been trying for 2 days to do something [18:38:15 CEST] <__jack__> marsupial: tell us [18:39:38 CEST] <marsupial> i am trying to record an HLS stream. i followed various guides , have located the m3u8 file with firebug. if i just try ffmpeg -i "stream.m3u8" , it says it cannot open the key [18:40:03 CEST] <__jack__> marsupial: you will need ffmpeg -i http://blabla/stream.m3u8, probably [18:40:05 CEST] <marsupial> so i downloaded the m3u8 file, and opened it in text.. then i downloaded the chunk file that was identified there.. and i found some URI IV key [18:40:13 CEST] <marsupial> yes i did that [18:40:23 CEST] <marsupial> i have the full url of the stream [18:40:51 CEST] <marsupial> this is the key information i was able to find [18:40:53 CEST] <marsupial> #EXT-X-KEY:METHOD=AES-128,URI="faxs://faxs.adobe.com",IV=0xc3585344ac1ecb2649fb4aa246614791 [18:41:14 CEST] <marsupial> but i have no idea how to make ffmpeg use that [18:41:23 CEST] <spoon> well if anyone thinks they can help with my RTP/RTSP decoding problem, let me know. I am willing to pay for the help. [18:42:07 CEST] <__jack__> marsupial: that is an crypted stream, with DRM, it will not work out of the box (because drm are made for that) [18:42:55 CEST] <marsupial> yes but i am a showtime subscriber so i should be able to do it [18:43:00 CEST] <marsupial> i know smarter people than me ARE doing it [18:43:08 CEST] <jkqxz> spoon: Perhaps if you describe in more detail what you are trying to do? [18:43:54 CEST] <spoon> I have written my own RTSP and RTP layers, now I need to decode incoming the video/audio [18:44:19 CEST] <spoon> I have the RTP pointer and size, and RTP payload pointer and size [18:44:26 CEST] <spoon> I have all the details that have been negotiated in RTSP [18:45:06 CEST] <spoon> I feel like I have all the information required, but actualy trying to decode the stream using FFMPEG is causing me headaches [18:45:32 CEST] <jkqxz> What is the format here, H.264? [18:45:34 CEST] <spoon> I have attempted to copy what FFMPEG is doing internally when it is in control of RTSP/RTP and I still get errors [18:45:41 CEST] <spoon> it's either H264, MJPEG or MPEG4 [18:47:06 CEST] <spoon> I am trying to get just H264 working at the moment, but the others must be supported too [18:47:09 CEST] <jkqxz> When you say you've decoded the RTP, you have already removed headers and dismantled STAPs and reassembled FUs into NAL units etc., or you want ffmpeg to do that bit for you? [18:47:27 CEST] <spoon> I have just read the RTP header [18:47:39 CEST] <spoon> I can pass the full RTP packet to FFMPEG if that's easier [18:48:25 CEST] <spoon> I have not done any of the things you have mentioned, so I guess yes, I would like FFMPEG to do that [18:48:49 CEST] <spoon> If it's easier/better to do it another way, that's fine too [18:55:08 CEST] <jkqxz> I don't think libavformat it is built to handle that sort of intermediate, but there may be some way I'm not aware of. If you dismantle the RTP packets yourself, then you can just feed the NAL units to libavcodec. [18:56:01 CEST] <spoon> ok thanks [18:57:00 CEST] <spoon> so I will need to dismantle the RTP payload further if it is H264 [18:57:59 CEST] <spoon> and I can pass the NAL units to avcodec_decode_video2? [18:59:04 CEST] <jkqxz> Yes. [18:59:52 CEST] <jkqxz> (Mashing frames together if they are more than one slice; the marker bit tells you whether a packet ends a frame.) [19:00:06 CEST] <spoon> wow that easy [19:00:15 CEST] <spoon> I can just literally join them together? [19:01:42 CEST] <jkqxz> The NAL units, yes. (After extracting them from the RTP stream.) [19:01:53 CEST] <spoon> ok awesome, I'm looking into now [19:02:03 CEST] <spoon> I'll probably have more questions once I've figured this out [19:02:12 CEST] <spoon> RTP packets can come out of order too, always fun! [19:02:29 CEST] <spoon> thanks for the help [19:04:54 CEST] <spoon> another rfc! https://tools.ietf.org/html/rfc6184#section-5.1 [19:07:00 CEST] <jkqxz> That's the one. Look at your stream first to see what types it actually has in it - I expect you need to be able to handle STAP-A and FU-A and can ignore the others, but do check. [19:08:32 CEST] <spoon> goodness I am really hoping FFMPEG can do this for me, I'll hunt around [19:08:52 CEST] <spoon> parsing this is a fairly big chunk of work [19:10:41 CEST] <jkqxz> The libavformat code to do it is in <http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavformat/rtpdec_h264.c;hb…>. [19:10:58 CEST] <spoon> thanks [19:11:05 CEST] <jkqxz> I don't see a way to access that externally, though there may be one I'm not aware of. [19:13:41 CEST] <jkqxz> (Copying the relevant bits of code out and hacking them would also work; it's all LGPL.) [19:14:09 CEST] <spoon> heh, possibly, all the params are very different [19:14:30 CEST] <spoon> At least I have something to work on, for too long I've just been confused [20:05:25 CEST] <dbugger> Hello guys [20:05:30 CEST] <dbugger> Guys and girls, of course [20:06:13 CEST] <dbugger> I have a small question: If I have 2 videos (300x100 pixels), how could I make a video that runs both of them at the same time, one on top of the other? (300x200) [20:17:24 CEST] <DHE> there's a vstack filter for pretty much this exact thing. [20:18:37 CEST] <DHE> $ ffmpeg -i inputfile1.mp4 -i inputfile2.mp4 -filter_complex '[0:v][1:v]vstack[videoout]' -map '[videoout]' -map '[0:a]' <codec options> output.mp4 [20:19:10 CEST] <dbugger> oh, let me try... [20:19:50 CEST] <DHE> so that would stack the videos and choose file1's audio [20:20:04 CEST] <dbugger> I actually want to mate it mute [20:20:48 CEST] <DHE> oh, then replace -map [0:a] with -an [20:20:59 CEST] <DHE> or actually just removing that entirely would be fine [20:21:54 CEST] <dbugger> im trying [20:23:31 CEST] <dbugger> Mmmm [20:23:32 CEST] <dbugger> it fails [20:23:41 CEST] <dbugger> but probably because I didnt give you the real dimensions of the video [20:23:56 CEST] <DHE> it shouldn't matter as long as they are identical [20:24:06 CEST] <dbugger> Im afraid they are not [20:24:15 CEST] <dbugger> Some cropping could be possible? [20:25:44 CEST] <DHE> To crop the first video, -filter_complex '[0:v]crop=w=1920:h=1080[cropout];[cropout][1:v]vstack[videoout]' [20:25:59 CEST] <dbugger> One single commando? [20:26:02 CEST] <dbugger> Or one after the other? [20:26:27 CEST] <DHE> it's a processing pipeline. the first video is cropped, then the cropped image is submitted as the first image to the vstack [20:26:42 CEST] <DHE> you can do quite a lot with one ffmpeg command if you're willing to write it out [20:26:49 CEST] <dbugger> so the final commando is... [20:27:23 CEST] <dbugger> mpeg -i inputfile1.mp4 -i inputfile2.mp4 -filter_complex '[0:v]crop=w=1920:h=1080[cropout];[cropout][1:v]vstack[videoout]' -filter_complex '[0:v][1:v]vstack[videoout]' -map '[videoout]' -map '[0:a]' output.mp4 [20:27:27 CEST] <dbugger> sorry [20:27:35 CEST] <dbugger> ffmpeg -i inputfile1.mp4 -i inputfile2.mp4 -filter_complex '[0:v]crop=w=1920:h=1080[cropout];[cropout][1:v]vstack[videoout]' -filter_complex '[0:v][1:v]vstack[videoout]' -map '[videoout]' -map '[0:a]' output.mp4 [20:27:38 CEST] <dbugger> Is that right? [20:28:23 CEST] <DHE> no, you've got two filter_complex commands, you still have the audio mapped in, and you should specify some codec options like "-c:v libx264 -b:v 3000k" or something [20:29:18 CEST] <dbugger> Then this? ffmpeg -i inputfile1.mp4 -i inputfile2.mp4 -filter_complex '[0:v]crop=w=1920:h=1080[cropout];[cropout][1:v]vstack[videoout]' '[0:v][1:v]vstack[videoout]' -map '[videoout]' -map '[0:a]' "-c:v libx264 -b:v 3000k" output.mp4 [20:29:39 CEST] <dbugger> I still forgot to take the audio out... but yeah [20:29:41 CEST] <dbugger> you know [20:30:39 CEST] <dbugger> is that one right? [20:33:13 CEST] <dbugger> I tried it and it gave me this error: [NULL @ 0x21a5640] Unable to find a suitable output format for '[0:v][1:v]vstack[videoout]' [20:33:13 CEST] <dbugger> [0:v][1:v]vstack[videoout]: Invalid argument [20:34:28 CEST] <DHE> I feel like you've never used the command-line very much... [20:35:26 CEST] <dbugger> I have used command line, just not much ffmpeg :P [20:35:39 CEST] <dbugger> oh wow, I added the quotes? [20:36:14 CEST] <DHE> and took the the "-filter_complex" but left in the actual filter that went with it [20:37:06 CEST] <dbugger> I corrected it: [20:37:10 CEST] <dbugger> ffmpeg -i v1.webm -i v2.webm -filter_complex '[1:v]crop=w=320:h=240[cropout];[cropout][1:v]vstack[videoout];[0:v][1:v]vstack[videoout]' -map '[videoout]' -c:v libx264 -b:v 3000k output.mp4 [20:37:13 CEST] <dbugger> That is now right, no? [20:38:06 CEST] <dbugger> oh damm [20:38:07 CEST] <dbugger> silly me [20:38:15 CEST] <dbugger> you were actually giving me before both filters together [20:38:28 CEST] <dbugger> Sorry, I thought they were separate, one for cropping and one for merging :P [20:39:22 CEST] <dbugger> Now it seems to be working [20:39:44 CEST] <dbugger> Oh, yeah! Excelent :) [20:39:50 CEST] <dbugger> Thanks mate! You are a saver! [20:44:14 CEST] <dbugger> DHE, what if I wanted to make more advanced layouts, instead of "one on top of the other"? Sort like "2 colums. 1 with 3 rows, and 1 with just 1 row"? [20:44:23 CEST] <dbugger> How could that be written? [20:51:08 CEST] <DHE> there's hstack and vstack so for full tiles you could assemble that. otherwise you'll have to get creative with the overlay filter [20:51:45 CEST] <dbugger> I see.... [20:51:52 CEST] <dbugger> So, lets see if I can come up with something.. [20:53:51 CEST] <dbugger> Interesting... so I have to make combinations then [20:55:46 CEST] <DHE> just off the top of my head, -filter_complex '[0:v][1:v][2:v]hstack=inputs=3[row1];[3:v][4:v][5:v]hstack=inputs=2[row2];[row1][row2]vstack[videoout]' # just made this up for a 3-wide, 2 high stack [20:56:07 CEST] <dbugger> ok... "inputs" seems useful [20:58:24 CEST] <l1l> is this the proper way to rip an mp3 ffmpeg -ss 19:40 -i Alok3.mp3 -t 5:00 aloktres.mp3 [20:58:39 CEST] <l1l> or should i use a different app [20:58:54 CEST] <dbugger> DHE, works great! [20:58:55 CEST] <dbugger> Thanks! [20:59:02 CEST] <dbugger> Now I just need to learn how to better crop images [20:59:08 CEST] <dbugger> But that can wait for latter [21:06:32 CEST] <ddmd> Could anyone please look at the difference between these two encodings? http://pastebin.com/Pgr3fjZQ One video displays whites as white and the other displays whites as grays in some browser's html video player. [21:07:57 CEST] <JEEB> try it in mpv and those are the actual colors of the video [21:08:14 CEST] <JEEB> on windows, http://mpv.srsfckn.biz/ [21:08:41 CEST] <JEEB> sometimes clips just don't have actual white in there and they are supposed to be grey. in other cases browsers just fail. [21:08:59 CEST] <JEEB> mpv's opengl renderer generally is the best bet at something getting it right [21:09:24 CEST] <JEEB> so if it looks grey in it, then either the content is supposed to be like that, or someone eff'd up [21:13:16 CEST] <ddmd> Thats the thing, it displays white on every native player I've tried and in every browser except for IE. So if it truly was off-white at source, why wouldn't the other players show it? Thanks for the suggestion I will try it in mpv. [21:14:23 CEST] <JEEB> many players also have various issues. both clips should be taken in as limited range according to that, although I recommend looking at the output of either `ffmpeg -i filename` or `ffprobe filename` [21:15:02 CEST] <JEEB> also mpv doesn't have a file-opening GUI so you basically can just drag and drop files on its binary [21:17:48 CEST] <Gues> Hey [21:18:22 CEST] <Gues> Does ffmpeg have an MP3 decoder? I don't see it here https://ffmpeg.org/ffmpeg-codecs.html#Audio-Decoders but ffmpeg tools seem to decode MP3s. [21:18:30 CEST] <JEEB> yes, it does [21:18:58 CEST] <Gues> Okay, ty [21:18:59 CEST] <JEEB> the manpages have a very small amount of things [21:19:57 CEST] <Gues> manpages? Are the contents related to web pages? [21:20:01 CEST] <JEEB> yes [21:20:07 CEST] <Gues> Ah [21:20:13 CEST] <JEEB> `ffmpeg -decoders` lists all of the decoders [21:20:20 CEST] <JEEB> in your FFmpeg [21:20:48 CEST] <Gues> ty [21:21:17 CEST] <Gues> I read on the website that ffmpeg uses lame to encode MP3. Does it implement its own decoder, or rely on a library? [21:21:29 CEST] <JEEB> the decoder is internal to libavcodec [21:21:47 CEST] <Gues> Okay. I was trying to determine if I should borther using ffmpeg or just its dependency [21:21:52 CEST] <Gues> *bother [21:22:17 CEST] <ddmd> The problematic video displays white correctly in mpv. I ran ffprobe on both files also. http://pastebin.com/x20qUUeC [21:22:53 CEST] <JEEB> ddmd: ok, then it's just a bug somewhere in the browser. enjoy reporting the issue [21:23:03 CEST] <c_14> Gues: https://ffmpeg.org/general.html#Supported-File-Formats_002c-Codecs-or-Featu… [21:23:32 CEST] <JEEB> the only weird part I could notice is that it has the BT.601 matrix and BT.709 primaries, but that shouldn't cause greyening [21:24:03 CEST] <JEEB> I guess IE might be misinterpreting the explicit color range tag, but I'd be surprised if that was the case :P [21:26:29 CEST] <ddmd> Its looking that way. I also cannot rule out it being an issue with a specific video card or hardware acceleration though. Thanks for your help. [21:36:50 CEST] <fp> When calculating psnr with my own tools (python/numpy) I get identical [21:36:50 CEST] <fp> results compared to ffmpeg and tiny_ssim if I look at the values per frame. [21:36:50 CEST] <fp> However, if I run ffmpeg || tiny_ssim an I look at the calculated average [21:36:50 CEST] <fp> I get something different. This I don't get! [21:36:58 CEST] <fp> see http://pastebin.com/AqUdkttH [21:56:55 CEST] <ATField> Can someone please help with a file-cutting problem? [21:56:56 CEST] <ATField> I am trying to cut a fragment from a movie file without re-encoding so that the result will be of the same quality as the original. IIUC, this will be the case if you start cutting from an I-frame so that ffmpeg wont have to decode anything to recover missing information. [21:56:58 CEST] <ATField> I am trying to find the I-frames timestamp by first generating a tile of all the nearby frames with their timestamps and their frame-type written on them (http://imgur.com/VvAefce, 3840×1800), and then using the needed I-frames timestamp as the startpoint and the TS of the frame right before the next I-frame as the endpoint. [21:56:59 CEST] <ATField> But the timestamp has to be wrong b\c when you play the file it starts off as a black screen until the player reaches an actual I-frame. Additionally, timestamps for the needed start- and end-points are different in Avidemux (09:56.346 -to 10:00.558 v.s. 09:56.012 -to 10:00.224). And since Avidemux seems to be able to cut the fragment out properly without having to re-encode... [21:57:01 CEST] <ATField> ...(both video... [21:57:02 CEST] <ATField> ...and audio outputs set to Copy), there should be a proper way of doing it that Im just missing. [21:57:04 CEST] <ATField> Ive also tried cutting by pkt_dts_time (http://stackoverflow.com/questions/14005110/) but again to no avail. [22:26:40 CEST] <SpeakerToMeat> What's the right way to reinterpret a 23.976 film to 24fps? copying video and audio tracks? [22:28:53 CEST] <JEEB> why would you reinterpret it? [22:29:06 CEST] <JEEB> the content is what it is, don't touch it without a reason [22:30:07 CEST] <JEEB> basically if you just change the timestamps on video you have to re-encode audio, if you actually start doing frame rate conversion then you'll be re-encoding video [22:30:17 CEST] <JEEB> neither of those alternatives sound especially good [22:30:48 CEST] <SpeakerToMeat> I don't want to reencode video, thus I need to tell it it's really 24 and not 23.976 [22:30:59 CEST] <JEEB> which means you will have to muck with the audio? [22:31:06 CEST] <JEEB> what are you thinking of achieving with this!? [22:31:09 CEST] <SpeakerToMeat> yuppers. yes. [22:31:26 CEST] <SpeakerToMeat> I'm thinking on achieving what I need to achieve. Are you saying it's not possible to do with ffmpeg? [22:31:59 CEST] <JEEB> well there's a resampler and a few other things so I'm pretty sure you can achieve what you want [22:32:04 CEST] <JEEB> but it just makes no sense to me [22:32:17 CEST] <JEEB> you're losing either video or audio quality for no perceivable benefit [22:32:43 CEST] <SpeakerToMeat> Well the benefit is being able to use it. not being able to use the video at all would be bad. [22:33:00 CEST] <SpeakerToMeat> Since I have a 23.976 video that needs to play in a device that will not do 23.976 [22:33:08 CEST] <JEEB> but it will do 24? [22:33:44 CEST] <SpeakerToMeat> Yes [22:34:14 CEST] <JEEB> given that 24000/1001 is the "broadcast" 24p that surprises me deeply. but hey, I have no idea what weird hardware you have there :) [22:35:12 CEST] <SpeakerToMeat> It's called a Digital Cinema media server, media block and projector [22:36:02 CEST] <JEEB> also, for the record I haven't found the way to redo the video timestamps with ffmpeg cli. I know that the API is capable of it but the thing doesn't seem to be available outside of the setpts video filter which requires re-encoding (which is usually something you want not to do) [22:36:06 CEST] <kepstin> ah, so an actual cinema projector. yeah, those are really 24p :) [22:36:21 CEST] <JEEB> I use L-SMASH's muxer generally to do the frame rate re-assumptions [22:36:27 CEST] <JEEB> and then ffmpeg for the rest [22:37:12 CEST] <JEEB> for audio there's the atempo filter [22:37:17 CEST] <SpeakerToMeat> kepstin: Yes. I can do the reinterpretation in premiere in another machine, since it'll have to go to dpx and jpeg2000 frames anyhow, but I'm subtitling it, and it's an easier workflow for me if I can subtitle in this machine first [22:38:14 CEST] <JEEB> http://ffmpeg.org/ffmpeg-all.html#atempo [22:38:35 CEST] <SpeakerToMeat> But my subtitle editor can't read dpx/jpeg2000 discrete files (afaik) to work with. And copying to the windows machine, reinterpreting it to 24, copying it back, doing the subtitles, copying it to premiere, burning the subtitles to image and then exporting is a pita [22:38:49 CEST] <SpeakerToMeat> thanks I'll check.. thouch I can sox the audio tracks if I have to. [22:38:51 CEST] <JEEB> I guess you feed 24 / (24000/1001) into that [22:39:15 CEST] <JEEB> btw, regarding subtitling - while jpeg2000 will be slow as hell to decode in any case, a recent enough aegisub should be able to read it [22:39:32 CEST] <JEEB> http://www.aegisub.org/ [22:39:51 CEST] <SpeakerToMeat> I work with subtitle edit, it's the best editor in the world for me. [22:39:57 CEST] <SpeakerToMeat> Exceptionally well done [22:40:00 CEST] <c_14> You can probably also use asetrate for o conversion of that scale [22:40:18 CEST] <JEEB> subtitle edit is something I'd mostly use for basics like OCR and editing [22:40:28 CEST] <JEEB> anything more requiring I'd probably pop aegisub for [22:40:38 CEST] <JEEB> also I think subtitle edit still bases on DirectShow, no? [22:40:46 CEST] <JEEB> (for the video input) [22:41:33 CEST] <kbarry_> What is the trick to seeing the "aliases" for commands in ffmpeg? [22:41:35 CEST] <JEEB> also man, aegisub hasn't gotten a release in a while :V [22:41:37 CEST] <kbarry_> Often i find help online, in the form of commands, but have the hardest time picking the solution apart, so I can learn/better understand what is actually happeneing [22:42:01 CEST] <c_14> That was fast [22:42:01 CEST] <JEEB> (it is under active development still, but man) [22:42:27 CEST] <SpeakerToMeat> JEEB: I use it for editing. Conversions, and retouching subtitle times, fixing sync, etc [22:42:34 CEST] <kbarry> for instance, I just found a link in the documentation that says "Control quality with -qscale:a (or the alias -q:a). " [22:43:29 CEST] <JEEB> SpeakerToMeat: whatever floats your boat, basically :) for me subtitle edit is something to do the very basics of ripping blu-ray or DVD subpictures, while further synch or styling or whatever would go into aegisub [22:44:51 CEST] <c_14> kbarry: besides just getting a feel for it? [22:45:18 CEST] <c_14> You could go through the source and create such a list [22:45:31 CEST] <c_14> I wouldn't recommend it though [22:45:55 CEST] <kbarry> So, [22:45:58 CEST] <JEEB> kbarry: and the documentation also wouldn't tell you that with some newer encoders q shouldn't be used [22:46:24 CEST] <kbarry> So, I am very new to ffmpeg, and I don't really have a more experiences mentor to rely on. [22:46:28 CEST] <c0rnw19> good evening [22:46:33 CEST] <kbarry> My main sources of info, are the docs, and here, and forums. [22:46:47 CEST] <kbarry> Just trying to get some basic tips on improving my resource access [22:46:59 CEST] <kbarry> IF the code is the best source of information, [22:47:27 CEST] <kbarry> How would I best "reverse engineer" the a command I find online? [22:47:42 CEST] <kbarry> IE, I am assuming there is some kind of built in help, [22:48:02 CEST] <kbarry> but I am not entirely sure I know how to use it. [22:48:05 CEST] <JEEB> ffmpeg-all.html which is all the manpages put together on ffmpeg.org is probably the best general docs you will find, although you shouldn't think from the explanations that something is the best way to do something [22:48:17 CEST] <JEEB> https://www.ffmpeg.org/ffmpeg-all.html [22:48:25 CEST] <JEEB> and when you find a parameter you can ctrl+F it [22:48:28 CEST] <c_14> If you want options for a specific encoder/muxer there's `ffmpeg -h encoder=' and `ffmpeg -h muxer=' [22:48:42 CEST] <c_14> Because not all options are in the manpages [22:48:47 CEST] <JEEB> that is true [22:49:08 CEST] <JEEB> also I love it how there's literally two video decoder entries in the manpage [22:49:11 CEST] <JEEB> hevc and rawvideo [22:49:25 CEST] <JEEB> both are recent enough that someone cared to write docs, I guess :) [22:49:40 CEST] <JEEB> allthough rawvideo as such is quite old already [22:57:14 CEST] <ATField> (Can my messages be seen or the registration issue makes them invisible?) [22:57:45 CEST] <durandal_1707> off topic [23:00:05 CEST] <kbarry> I still find the documentation hard to grasp. [23:00:59 CEST] <kbarry> I think its a problem with, the syntax can vary so widely (ffmpeg is so versetile), that maybe its just hard to find an example that succinctly covers a topic. [23:01:45 CEST] <durandal_1707> what's your goal? [23:02:45 CEST] <iive> ATField: we see you. ask your question [23:03:40 CEST] <kbarry> Idurandal_1707: is your last directed atme? [23:04:04 CEST] <durandal_1707> yes [23:04:08 CEST] <ATField> iive: Thanks, I thought I was being rendered invisible by the network. Ive asked it higher, will post a superuser question link here soon instead. [23:05:47 CEST] <kbarry> My goal is to get better at helping myself. I recognize a lack of ability to help myself to the documentation, or even how to find help while using the CLI. [23:05:56 CEST] <iive> ATField: your question above is also visible. [23:06:07 CEST] <kbarry> I want to be able to do more, myself, before I come in here asking for help. [23:06:19 CEST] <ATField> iive: Thanks again. [23:06:31 CEST] <kbarry> Trying to LEvel up to "Level II Newb" [23:08:41 CEST] <iive> ATField: one small hint. you can put -ss at the -i input demuxer set of options and it would be seeking to a keyframe first... [23:09:04 CEST] <iive> if i remember correctly it's -ss first then -i ... i don't use it... [23:16:12 CEST] <ATField> iive: Is this syntax correct? "ffmpeg -ss 00:09:56.012 -i "INPUT.mkv" -map 0:0 -map 0:4 -map 0:5 -map 0:6 -t 4.212 -vcodec copy -acodec copy test09.mkv" Still doesnt produce a viable result (the 4.212 is 10:00.224 09:56.012). [23:23:17 CEST] <SpeakerToMeat> Ok what am I effing up? ffmpeg -i file.mov -vc copy setpts="N/(24*TB)" -ac copy atempo="24/(24000/1001)" file-24.mov [23:24:01 CEST] <furq> SpeakerToMeat: -c:v not -vc [23:24:15 CEST] <furq> and -vf setpts [23:24:23 CEST] <c_14> and -af atempo [23:24:27 CEST] <SpeakerToMeat> Other than not knowing if -ac copy will work with atempo [23:24:29 CEST] <SpeakerToMeat> thanks [23:24:32 CEST] <c_14> It won't [23:24:32 CEST] <SpeakerToMeat> of course, filters [23:24:36 CEST] <furq> and yeah -vf and -af require reencoding [23:24:57 CEST] <SpeakerToMeat> furq: So there's no way to apply setpts without reencoding? [23:25:34 CEST] <furq> no [23:25:50 CEST] <SpeakerToMeat> Then I'll end up doing this in premiere [23:26:11 CEST] <furq> there are convoluted ways to change the framerate of a video without reencoding but i've never had cause to use them [23:26:21 CEST] <furq> someone else will probably remember better than i do [23:27:03 CEST] <furq> you'll definitely need to reencode for atempo though [23:35:58 CEST] <iive> ATField: i don't know. you seem to use timecode notation, but I don't know if ffmpeg actually supports it. [23:36:24 CEST] <iive> aka afair .012 should mean 12'th frame, but ffmpeg might get it as 0.012 second [23:37:55 CEST] <ATField> iive: The manual says two different time unit formats can be used, sexagesimal (HOURS:MM:SS.MICROSECONDS, as in 01:23:45.678), or in seconds. And the format works with everything else, or even for this very task if I dont use direct-copy instead of re-encoding. [23:40:10 CEST] <iive> timecode is a thing, like a standard. not just any time code. [23:40:17 CEST] <iive> just wanted to make sure... [23:41:07 CEST] <iive> btw, don't forget to take a look of the manual `man ffmpeg` [23:42:56 CEST] <ATField> The question on SU with some more code, if anyones interested to take a look: http://superuser.com/questions/1076283/how-to-cut-starting-precisely-from-a… [23:43:40 CEST] <ATField> iive: Yeah, it usually helps to check and make sure, thanks. [00:00:00 CEST] --- Fri May 13 2016

1 0

ffmpeg-devel.log.20160511
by burek 12 May '16

12 May '16

[00:43:18 CEST] <cone-640> ffmpeg 03Michael Niedermayer 07master:8efaee3710ba: avformat/oggparseopus: Check that granule pos is within the supported range [00:43:19 CEST] <cone-640> ffmpeg 03Michael Niedermayer 07master:ea791c080dd5: avformat/utils: Check bps before using it in a shift in ff_get_pcm_codec_id() [01:03:52 CEST] <kierank> jamrial: oddly, no [01:12:21 CEST] <jamrial> kierank: ok, going to push it then. thanks [01:32:30 CEST] <cone-640> ffmpeg 03Felt, Patrick 07master:fce75131229b: avdevice/decklink_dec: Convert decklink input module to use codecpar [01:34:49 CEST] <cone-640> ffmpeg 03foo86 07master:6c44696b3d50: avcodec/dca: add DTS Express (LBR) decoder [03:12:08 CEST] <cone-640> ffmpeg 03James Almer 07master:b2244fa0a624: avcodec/rscc: check input buffer size for deflate mode [03:17:15 CEST] <cone-640> ffmpeg 03James Almer 07release/3.0:8dce66d33d41: avcodec/rscc: check input buffer size for deflate mode [03:34:01 CEST] <cone-640> ffmpeg 03Rodger Combs 07master:d645182227e8: lavfi/drawutils: support NV12 and NV21 [04:22:12 CEST] <cone-640> ffmpeg 03Chris Cunningham 07master:542f725964e5: libavformat/oggdec: Free stream private when header parsing fails. [04:36:31 CEST] <cone-640> ffmpeg 03James Almer 07master:c8c14d0ffc51: aarch64/synth_filter: fix compilation [09:39:36 CEST] <BtbN> Anyone with knowledge about h264, what could h264Config.outputBufferingPeriodSEI/outputPictureTimingSEI do? I'm getting some reports that enabling those causes propper CBR padding, which doesn't make much sense to me. [09:40:06 CEST] <nevcairiel> it should just emit the SEIs, not change anything else [09:40:28 CEST] <nevcairiel> or so the option names sound [09:41:34 CEST] <BtbN> Yeah, that's what I think. But aparently, without it, in CBR mode, it causes CBR padding. [09:41:53 CEST] <BtbN> Or disables CBR padding, rather. [09:42:04 CEST] <BtbN> Is there any downside to just enabling it? [09:42:20 CEST] <BtbN> Should just be a few extra fields in the bitstream? [10:14:55 CEST] <nevcairiel> maybe it thinks to itself that without these SEIs its not a fully compliant CBR bitstream either way, so no point producing the CBR padding which is just overhead [10:15:29 CEST] <BtbN> So can I safely just enable those flags, or do they cause issues in some modes? [10:26:46 CEST] <nevcairiel> its probably fine if running in CBR, but who knows if something might barf on that [10:52:44 CEST] <cone-199> ffmpeg 03Timo Rothenpieler 07master:3a9df7dfec89: avcodec/nvenc: Generate bufferingPeriod/pictureTiming SEI [11:47:08 CEST] <atomnuker> michaelni: how come av_crc_init() wants a polynomial with the highest power stripped yet with !CONFIG_HARDCODED_TABLES av_crc_get_table() uses the standard representation? [12:00:35 CEST] <kierank> atomnuker: are you sure it's using the standard representation [12:02:14 CEST] <atomnuker> [AV_CRC_32_IEEE] = { 0, 32, 0x04C11DB7 }, the polynomial is as it's defined [12:03:15 CEST] <kierank> there's a leading 1 iirc [12:03:49 CEST] <kierank> the polynomial is x32 + x26 + x23 + x22 + x16 + x12 + x11 + x10 + x8 + x7 + x5 + x4 + x2 + x + 1. [12:06:02 CEST] <atomnuker> huh, that x^32 power is missing from 0x04C11DB7 [12:06:35 CEST] <atomnuker> it's correct then [12:07:07 CEST] <kierank> yes that's the point [12:42:57 CEST] <michaelni> atomnuker, if it didnt then it would not be possible to use polynomes as large as the type (with uint32 31bit would be max, with uint64, 63 would be max) [12:47:19 CEST] <atomnuker> michaelni: think it's possible to use av_crc with non-standard widths, like 18 bits? [12:48:11 CEST] <atomnuker> because the polynomial I have is 18 bits [12:48:57 CEST] <atomnuker> and that for (c = i, j = 0; j < 8; j++) loop in av_crc_init makes me suspicious that it assumes a power of two #bits for the polynomial [12:54:05 CEST] <michaelni> atomnuker, it should work with non mod 8 bits, but someone needs to test if it actually does work, i just dont see why it wouldnt [12:55:48 CEST] <michaelni> if you find a case that doesnt work, please tell me / provide a testcase that fails, im interrested to see / know [12:59:11 CEST] <wm4> nevcairiel: do you think we could use SRW locks on Windows to get non-leaking, statically initializable mutexes? [13:00:45 CEST] <cone-199> ffmpeg 03Timo Rothenpieler 07master:c921ca9b5d27: avcodec/nvenc: Write buffPeriod/picTime SEI in both CBR modes [16:13:52 CEST] <jamrial> nevcairiel: msvc doesn't like a struct in dca_lbr, can you look at it? [16:15:17 CEST] <nevcairiel> this type of inline declaration and init is probably not supported [16:15:32 CEST] <nevcairiel> looks odd from a style perspective as well [16:16:13 CEST] <nevcairiel> i can try to rewrite it a bit later [16:19:19 CEST] <cone-199> ffmpeg 03Diego Biurrun 07master:330177b50842: build: Group declarations for hw-accelerated de-/encoding separately [16:19:19 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:ce3037ac8ee7: Merge commit '330177b508420a553083df94f22cbd5142de0f4a' [16:19:43 CEST] <jamrial> nevcairiel: ok, thanks [16:22:27 CEST] <jamrial> speaking of which, tried to write fate tests for them using the samples foo86 provided, and the difference between x87 and sse math was pretty big [16:22:32 CEST] <jamrial> creating the ref pcm file with x87 then running fate with an sse build required a fuzz >250 to pass for the 5.1 sample [16:22:59 CEST] <jamrial> whereas a fuzz of 1 was enough the mono and stereo samples [16:23:05 CEST] <nevcairiel> that doesnt sound r ight [16:23:21 CEST] <jamrial> lfe_iir_c() in dcadsp.c is at fault [16:26:23 CEST] <nevcairiel> hm odd the msdn docs list such a type of declaration as valid [16:26:31 CEST] <jamrial> nevcairiel: http://pastebin.com/U2WqhmGA see if you can reproduce it whenever you have time [16:27:05 CEST] <jamrial> try initialialize the struct with { 0 } instead of { }? [16:27:11 CEST] <nevcairiel> thats the first thing itried [16:27:52 CEST] <nevcairiel> even removing the init still fails [16:28:22 CEST] <nevcairiel> .. what doesnt it like [16:30:36 CEST] <nevcairiel> oh i compiled and edited different files [16:31:36 CEST] <nevcairiel> yeah changing the init to 0 helps [16:32:11 CEST] <Daemon404> merging this test split hting is such tedium... [16:32:58 CEST] <jamrial> Daemon404: as i said i can help you with the modules they don't have if you want [16:35:11 CEST] <Daemon404> jamrial, sure can you work on that now? [16:35:21 CEST] <Daemon404> im current going through and doing the ones that we do have [16:35:26 CEST] <Daemon404> currently* [16:35:32 CEST] <jamrial> yeah [16:35:56 CEST] <cone-199> ffmpeg 03Hendrik Leppkes 07master:27506aceda81: avcodec/dca_lbr: explicitly initialize structs with zero [16:36:17 CEST] <jamrial> create a branch in your github with your current work and i'll push there like we did for codecpar [16:38:08 CEST] <Daemon404> jamrial, i dont have a branch [16:38:10 CEST] <Daemon404> im mid merge [16:38:20 CEST] <jamrial> ah [16:38:42 CEST] <Daemon404> i dont have a list of stuff we do have but they dont [16:38:45 CEST] <Daemon404> that would be helpful [16:41:25 CEST] <jamrial> Daemon404: http://pastebin.com/d1SNA1vd [16:42:21 CEST] <Daemon404> ... that is a lot [16:42:37 CEST] <Daemon404> can you start work? you dont need a branch [16:42:44 CEST] <Daemon404> just cp file.c file-test.c [16:42:47 CEST] <Daemon404> and edit the two [16:42:57 CEST] <Daemon404> the grunt work is what takes the longest [16:46:21 CEST] <jamrial> alright [17:29:09 CEST] <Daemon404> jamrial, finished the ones in the merge [17:29:14 CEST] <Daemon404> shall i push it to a branch for you [17:29:39 CEST] <jamrial> yeah [17:29:54 CEST] <jamrial> i'm also done with libavutil modules [17:30:23 CEST] <Daemon404> https://github.com/dwbuiten/FFmpeg/tree/testprogs [17:30:33 CEST] <Daemon404> please poke me when you are done [17:30:54 CEST] <jamrial> looks like there's also some libavcodec testprogs we do that they don't [18:10:45 CEST] <wm4> ubitux: more subtitle crap https://trac.ffmpeg.org/ticket/5539 [18:11:04 CEST] <ubitux> yep i had this one or a similar [18:11:27 CEST] <ubitux> in the huge db of subs [18:11:45 CEST] <ubitux> i saw some with just "->" too [18:39:43 CEST] <jamrial> Daemon404: pushed [18:47:57 CEST] <Daemon404> jamrial, ok cool [18:48:01 CEST] <Daemon404> ill check after dinner [19:40:00 CEST] <atomnuker> ping Gramner, look at PM whenever you have the time [19:40:29 CEST] <kierank> Meh just post it in the channel [19:40:42 CEST] <kierank> How can we make slow crc fast [19:40:51 CEST] <atomnuker> https://0x0.st/NhF.c [19:40:56 CEST] <atomnuker> the test program [19:41:02 CEST] <atomnuker> ask me if someone needs a test file [19:41:41 CEST] <iive> the kernel have a number of fast crc implementation, including ones using special instructions [19:45:36 CEST] <kierank> slow crc is slow :( [19:46:57 CEST] <jamrial> kierank: if you only need the standard crc32 to be fast, you could try to write an asm version using the sse4 crc32 instruction [19:47:16 CEST] <kierank> Nope a special crc [19:47:29 CEST] <jamrial> then dunno :p [19:48:00 CEST] <kierank> atomnuker: https://matt.sh/redis-crcspeed [19:48:11 CEST] <kierank> Looks similar to michaelni's approach [19:48:12 CEST] <jamrial> make sure to set CRC_TABLE_SIZE to 1024 if you're using a custom crc that's no defined in crc.h [19:51:10 CEST] <atomnuker> yeah, that's michael's approach basically, extended to 32 bits [19:52:13 CEST] <kierank> Might be doable for us except the interleaving hurts [19:52:31 CEST] <atomnuker> sadly our crc is 18 bits wide and our data is 10, so can't fit 2 into the CRC [19:53:58 CEST] <kierank> Might have to yasm it then for our big loop [19:54:01 CEST] <kierank> Eugh [19:54:55 CEST] <atomnuker> wouldn't work that well since we'd have to carry it over in a separate register between lines [19:55:45 CEST] <Daemon404> jamrial, ok looking [20:07:06 CEST] <Daemon404> jamrial, seems good [20:07:08 CEST] <Daemon404> fate builds and passes [20:07:53 CEST] <Daemon404> just doing a readthrough for e.g. copyright headers [20:10:33 CEST] <jamrial> already fixed that in one of the commits i pushed [20:11:06 CEST] <jamrial> at least those covered by fate-source [20:11:30 CEST] <Daemon404> yeah i see that [20:11:37 CEST] <rcombs> what's this CRC for anyway [20:11:42 CEST] <rcombs> who uses an 18-bit CRC [20:13:41 CEST] <cone-199> ffmpeg 03Diego Biurrun 07master:d12b5b2f135a: build: Split test programs off into separate files [20:13:43 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:96d616052b3d: Merge commit 'd12b5b2f135aade4099f4b26b0fe678656158c13' [20:14:49 CEST] <cone-199> ffmpeg 03Martin Storsjö 07master:798845ce7e5b: testprogs: Add missing libm.h includes [20:14:50 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:4f173d055c3a: Merge commit '798845ce7e5b7fdd17c7269c5d267fb487d9c46f' [20:15:31 CEST] <cone-199> ffmpeg 03Martin Storsjö 07master:b8e899f4bf5f: mmaldec: Use imgutils.h for copying frames [20:15:32 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:f4c88eef95e6: Merge commit 'b8e899f4bf5f09900aa71552112d32a5566b6baf' [20:17:07 CEST] <kierank> rcombs: hardware [20:19:56 CEST] <cone-199> ffmpeg 03Martin Storsjö 07master:e8919ec486a5: libavcodec: Add H264/MPEG4 encoders based on OpenMAX IL [20:19:56 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:0e3872322082: Merge commit 'e8919ec486a5559fdcf366e347be0656d904a87f' [20:23:41 CEST] <cone-199> ffmpeg 03Martin Storsjö 07master:f1cd9b03f3fa: omx: Add support for broadcom OMX on raspberry pi [20:23:42 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:e330ab0fb7cb: Merge commit 'f1cd9b03f3fa875eb5e394281b4b688cec611658' [20:24:40 CEST] <cone-199> ffmpeg 03Martin Storsjö 07master:1bb56abb9b37: omx: Add support for zerocopy input of frames [20:24:41 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:ec1d8abfb9bf: Merge commit '1bb56abb9b37bd208a66164339c92cad59b1087b' [20:26:40 CEST] <cone-199> ffmpeg 03Martin Storsjö 07master:9d4d9be538fa: libavcodec: Document that encoders may use the framerate field in AVCodecContext [20:26:42 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:75a1fc0376cc: Merge commit '9d4d9be538faa537440fff37d3b7ecf322911a55' [20:39:28 CEST] <cone-199> ffmpeg 03Anton Khirnov 07master:d7abe900c3d3: FATE: add an H.264 test with invalid reference lists [20:39:29 CEST] <cone-199> ffmpeg 03Anton Khirnov 07master:18019f8cb966: FATE: add an H.264 test with unescaped extradata [20:39:30 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:bc3d2f25783a: Merge commit '18019f8cb9663dd1032c65aa453eaec18d641905' [20:41:08 CEST] <cone-199> ffmpeg 03Mark Thompson 07master:f6b85523692b: vaapi_encode: Refactor slightly to allow easier setting of global options [20:41:09 CEST] <cone-199> ffmpeg 03Mark Thompson 07master:6e8f66fc932b: vaapi_h264: Add constant-bitrate encode support [20:41:10 CEST] <cone-199> ffmpeg 03Mark Thompson 07master:69b06ed42809: vaapi_encode: Add support for codec-local options [20:41:11 CEST] <cone-199> ffmpeg 03Mark Thompson 07master:9629701ce9de: vaapi_h264: Add -qp option, use it to replace use of -global_quality [20:41:12 CEST] <cone-199> ffmpeg 03Mark Thompson 07master:fcf536b13014: vaapi_h264: Add encode quality option (for quality-speed tradeoff) [20:41:13 CEST] <cone-199> ffmpeg 03Mark Thompson 07master:f70e4627933b: vaapi_h265: Add constant-bitrate encode support [20:41:14 CEST] <cone-199> ffmpeg 03Mark Thompson 07master:92fdea37477b: vaapi_h265: Add -qp option, use it to replace use of -global_quality [20:41:15 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:8ed427f9f92f: Merge commit '92fdea37477b5a2d1329e5ef0773e24473fa8f12' [20:41:19 CEST] <Daemon404> <_<; [20:41:27 CEST] <JEEB> cone decided to spam [20:42:58 CEST] <BBB> do we have a avfilter that gives me the dc per frame? [20:43:10 CEST] <BBB> like, I want to detect dark and light scenes in a particular video [20:45:32 CEST] <cone-199> ffmpeg 03Anton Khirnov 07master:5e1a3ea3ba7b: lavc: move the vaapi encoders further down in the list of codecs [20:45:34 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:ed8bacfb5ca8: Merge commit '5e1a3ea3ba7bb0c71d931e93e60fb75f51b0cc1a' [20:46:35 CEST] <cone-199> ffmpeg 03Anton Khirnov 07master:a0f469da744d: hwcontext: initialize sw_format in av_hwframe_ctx_alloc() [20:46:37 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:c71c3b9ff4f1: Merge commit 'a0f469da744db83db32f3fe13186ee4aa2bc7dc5' [20:46:41 CEST] <durandal_1707> BBB: dc is ? [20:47:17 CEST] <Daemon404> dc coefficient [20:47:18 CEST] <Daemon404> i assume [20:51:04 CEST] <BBB> yes [20:51:08 CEST] <BBB> but of the whole image, not 1 block [20:51:18 CEST] <BBB> average of dc coefficients over whole image is fine also [20:52:05 CEST] <durandal_1707> still dc is what? [20:53:27 CEST] <Daemon404> of the dct of the whole image [20:55:00 CEST] <durandal_1707> not done.. I guess Dct stuff is only in lavc [20:55:19 CEST] <kurosu_> why go through a dct, it's just average of pixels * constant [20:55:50 CEST] <fritsch> thought exactly the same [20:57:03 CEST] <kurosu_> isn't there an histogram filter or pixel stats or some such, with some post processing that might do it [20:57:59 CEST] <kurosu_> but histogram is overkill compared to psadb + paddd over an image [21:09:40 CEST] <cone-199> ffmpeg 03Michael Niedermayer 07master:bf29794022db: avcodec/dca_lbr: Fix "warning: missing braces around initializer" [21:09:40 CEST] <cone-199> ffmpeg 03Christophe Gisquet 07master:9ca16bdd3f04: lossless audio dsp: unroll [21:23:55 CEST] <michaelni> Daemon404, "make testprogs" is broken [21:24:26 CEST] <michaelni> make: *** [libavfilter/drawutils-test] Error 1 [21:24:27 CEST] <michaelni> make: *** [libavcodec/avfft-test.o] Error 1 [21:28:05 CEST] <Daemon404> checking [21:28:36 CEST] <michaelni> avfft is just a missing include mem.h [21:28:50 CEST] <Daemon404> michaelni, does fate not build these programs? [21:28:59 CEST] <michaelni> apparently not [21:29:03 CEST] <michaelni> make fate passed here [21:29:06 CEST] <Daemon404> yeah same [21:29:12 CEST] <Daemon404> i thought fate would have built them all [21:29:14 CEST] <Daemon404> guess not... [21:33:05 CEST] <michaelni> "make check" should run "all alltools examples testprogs fate" [21:33:29 CEST] <Daemon404> was not aware of that before [21:37:06 CEST] <jamrial> drawutils should be added to fate, i guess [21:37:45 CEST] <jamrial> like every other testprog [21:38:17 CEST] <Daemon404> jamrial, looks like it wasnt converted [21:38:24 CEST] <Daemon404> because we didnt see it in fate [21:38:37 CEST] <jamrial> yeah [21:39:35 CEST] <jamrial> i wonder why it's part of testprogs but was never added to fate, like avfilter's formats [21:41:34 CEST] <Daemon404> no idea [21:44:12 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:64fbf2f03e4e: avfft-test: Add missing mem.h include [21:44:13 CEST] <cone-199> ffmpeg 03Derek Buitenhuis 07master:f693184557e4: Split drawutils test out into separate file [21:44:15 CEST] <Daemon404> michaelni, fixed [21:44:31 CEST] <Daemon404> testprogs works [21:44:57 CEST] <michaelni> thx [22:05:26 CEST] <mateo`> sorry for the mistake in my email, the neon version of the resampling function is actually faster. [22:27:05 CEST] <fritsch> https://bugzilla.gnome.org/show_bug.cgi?id=766050#c5 <- gb is alive - nice :-) [22:29:32 CEST] <wm4> just not interested in ffmpeg anymore [22:30:13 CEST] <fritsch> don't think so ... [00:00:00 CEST] --- Thu May 12 2016

1 0

ffmpeg.log.20160511
by burek 12 May '16

12 May '16

[00:30:03 CEST] <vade> im transcoding a video , and I have the same number of packets and frames from source and destination, but my resulting video stream (the only stream) ha a duration thats slightly off. my source is 5 seconds on the dot, my resulting stream is 4.96. Both have 120 frames. original is 24 fps, my resulting is 24.20 which is the difference in duration. is there a trick to marking my output context duration / timebase to force absolute frame rate? sh [00:30:04 CEST] <vade> I NOT be setting certain codec contex flags on encode, like average frame rate, or stuff like that? [01:03:11 CEST] <esdwdftty> Big or not big advantage to decoding video (video players) between 64 bit and 32 bit OS? In encoding the difference I have on my CPU is ~ 40% faster (64 Bit). When decoding the difference I do not see significant during playback of a video file the CPU load is about the same average 60% of CPU 720p 30 frame. The encoding I do rarely. [01:04:26 CEST] <vade> youre probably waiting to decode most of the time for realtime playback I imagine. You can likeply playback orders of magnitude faster than realtime. [01:04:58 CEST] <vade> i bet a profile is just sleeping / spinning waiting for an appropriate time to get the next frame, thus similar CPU. Just a guess. [01:13:03 CEST] <yongyung> What is the difference between using (libx264) -b:v 1000k -bufsize 1000k and -maxrate 1000k -bufsize 1000k? [01:15:18 CEST] <esdwdftty> The translation software. Me here that interests it makes sense to change OS Windows 32 bit on Linux 64 bit only for video decoding it CPU. I have Windows 32 bit can install Linux 64 BIt. Please write without fanaticism for Linux or dislike for Windows. Linux OS difficult for me if something it is to do yourself. [01:19:20 CEST] <neuro_sys> I'm tinkering with libav*. After av_read_frame(), if I call avcodec_decode_video2 for video stream, av_read_frame never finishes, and main loop never breaks. Do we have to always use filters (as the example I'm following does)? [01:22:44 CEST] <esdwdftty> I did not see the benefits when watching video files (decoding on CPU) between 64 bit Linux and 32 bit Windows. [01:23:02 CEST] <esdwdftty> dxva i see [01:23:07 CEST] <neuro_sys> https://gist.github.com/neuro-sys/4a3897e115b4969b35277c44c85fd8d6 [01:25:54 CEST] <esdwdftty> Higher frequency of the processor too, the difference is visible. [01:31:56 CEST] <esdwdftty> Humor: maybe can need compare Windows 2000 16 bit with OS 64 bit then the difference will show itself. [03:17:19 CEST] <cobadger> baptiste: pm? [03:59:47 CEST] <baptiste> cobadger, shoot [04:02:09 CEST] <cobadger> baptiste: opened pm to you [04:53:11 CEST] <h64> hello. i'm having a weird problem with -map_metadata 0, the creation_date is off by a few hours. [04:53:18 CEST] <h64> nearly a day [04:53:45 CEST] <h64> in other words, it's not directly copying the creation date [04:54:32 CEST] <h64> maybe it's losing timezone information? input is UTV [04:54:34 CEST] <h64> *UTC [04:55:43 CEST] <h64> result is ahead by exactly 6 hours [04:56:36 CEST] <h64> (sorry i thought it was nearly a day because the days werent the same and i am bad at math lol) [05:05:34 CEST] <h64> ok i got it fixed. i set environment variable to TZ=UTC and no conversion was done. [05:06:10 CEST] <h64> perhaps theres a timezone conversion bug in ffmpeg? [07:02:26 CEST] <yongyung> Are you guys aware of any proper ways to upload videos to YT? With that I mean resumable uploads, bandwidth caps, error detection for individual chunks (I know tcp does that but it's not that great as I had to find out when I had to re-upload a video I had uploaded over night). [07:30:31 CEST] <furq> yongyung: https://github.com/youtube/api-samples/blob/master/python/upload_video.py appears to support resumption [07:30:38 CEST] <furq> no idea about the other two though [07:32:56 CEST] Action: davidshen84 [07:32:59 CEST] <davidshen84> hello [07:33:29 CEST] <davidshen84> I am new to ffmpeg library. I have some experience using the ffmpeg command line tool [07:33:38 CEST] <davidshen84> anybody? [07:34:04 CEST] <davidshen84> I want to create a simple python tool to read the video stream information [07:35:42 CEST] <davidshen84> e.g. Stream #0:1(eng): Video: vc1 (Advanced) (WVC1 / 0x31435657), yuv420p, 1280x720, 5942 kb/s, 29.97 fps, 29.97 tbr, 1k tbn [07:35:50 CEST] <davidshen84> any idea how to do that? [09:40:32 CEST] <Anaphaxeton> goodmorning [09:41:00 CEST] <Anaphaxeton> there is a video that i want to transcode to embed subtitles to [09:41:28 CEST] <Anaphaxeton> it is a hight quality h264 etc etc video [09:41:59 CEST] <Anaphaxeton> i don't want to downgrade the quality and i am willling to make a sacrifice on size [09:43:46 CEST] <Anaphaxeton> i thought about encoding it into mpeg2, because it is very simple compared to h264 and i dont want a psychovisual encoding over psychovisually encoded video [09:43:53 CEST] <Anaphaxeton> what is your opinion? [09:45:04 CEST] <bmariesan> Hi everyone, I'm experiencing some issues with ffmpeg udp streaming to a remote device. I have either two computers in different networks or one computer and one mobile device. Since they are in different networks I'm utilising a NAT traversal technique called UDP hole punching, meaning that I'm setting an initial connection to a HTTP server from both devices and the server finds out the public IPs and the ports opened in the NAT for [09:45:13 CEST] <furq> i don't think psy will really make much of a difference but you could just use x264 and turn psy off if you're really concerned [09:45:28 CEST] <bmariesan> You can find my ffmpeg command and ffplay commands at http://pastebin.com/QbTcgswX [09:48:26 CEST] <Anaphaxeton> furq, in principle, what should one do for encoding in a case like mine? [09:50:18 CEST] <furq> there isn't any magic way to not lose any visual quality short of encoding lossless [09:50:28 CEST] <furq> personally i'd just use x264 with a low crf [09:50:51 CEST] <Anaphaxeton> ok! much appreciated :) [10:02:47 CEST] <Anaphaxeton> i like the subtitles vlc puts out a lot. how could i achieve a similar quality when embedding subs to the video? [10:29:03 CEST] <adamk__> hey, does anyone know what might cause this error? Invalid option or argument: 'ref=6', parsed as 'ref' = '6' [10:29:25 CEST] <adamk__> i thought ref was an option [10:30:16 CEST] <furq> -refs [10:32:14 CEST] <adamk__> furq: thanks, this seems to work. i was looking at another file which was encoded and it had that in its encoding settings [10:34:37 CEST] <furq> -ref is the x264 option [12:22:00 CEST] <neuro_sys> reading from a sequence of pngs versus a h264 video to encode a video in terms of speeds didn't make any difference. I'd expect a difference at the least. Both took 50 seconds. [12:24:12 CEST] <neuro_sys> oh reading from BMPs now, seems faster [12:24:49 CEST] <neuro_sys> only slightly [12:26:16 CEST] <mr_lou> So I've noticed, that when I create videoclips in my editor, I'm always doing the exact same thing: Simply putting each clip right after each other with a simple crossfade effect. There has to be an easier way to do this without using a video-editor? An ffmpeg command to simply take a bunch of mp4 files and mix them together, crossfading the first/last second of each? [12:27:22 CEST] <neuro_sys> yeah, sounds like it can easily be automated [12:37:32 CEST] <mr_lou> Yea I think it must. Googling isn't bringing me any examples though. [12:40:04 CEST] <mr_lou> I can find examples that require me to know the durations of each video. [12:40:10 CEST] <mr_lou> That's not good enough. :-) [14:19:40 CEST] <neuro_sys> I need some good reading on PTS to understand it [14:19:56 CEST] <DHE> neuro_sys: PNGs are compressed with zlib, so while timing being identical is a bit surprising, being slow is expected [14:26:47 CEST] <neuro_sys> fair enough [14:29:07 CEST] <n1cksn1ck> hi. i am trying to rotate a video file from landscape to Portrait? But cant seem to find the correct Adv options filter command. Can anyone help me rotate/re-orientate my video file or know the correct Adv options command? Any help will much apprecieate : ) [14:34:44 CEST] <furq> n1cksn1ck: https://ffmpeg.org/ffmpeg-filters.html#rotate [14:35:03 CEST] <furq> or you can do it without reencoding using mp4 metadata (if it's an mp4) [14:36:27 CEST] <furq> -c copy -metadata:s:v:0 rotate=90 [14:36:38 CEST] <furq> that depends on player support though [14:40:45 CEST] <n1cksn1ck> hi furq. thanks for your suggestion. I have been to that page. I didnt try the rotate command. It seemed complicated. I did try the transpose command but got an error. [14:42:58 CEST] <n1cksn1ck> i will try the command as you have suggested. And let you know how it went. [14:47:16 CEST] <n1cksn1ck> furq: Wow. It accepted the command. So... so far so good. When the process is finished I will let you know if ive been successful(in rotating the ouput). Thanks heaps furq : ) [14:48:21 CEST] <furq> that should be pretty much instant [14:50:45 CEST] <n1cksn1ck> not sure what you mean there furq. Its a 400mb file. I do know that it is approx 6000 frames. And it is only at 1300. I dont have new/modern pc. It's about 10 yrs pld. [14:51:48 CEST] <furq> if you're just changing the metadata then it shouldn't be encoding anything [14:54:49 CEST] <n1cksn1ck> furq: um. im not just doing a rotation. im also hoping to resize it. Its a video my aunty took on her ph. But she took it in landscape. She asked if I could turn it the right way up. Some how, she managed to take a video 1920x1080. [14:55:17 CEST] <furq> it would be faster to just rotate it [14:56:01 CEST] <n1cksn1ck> oh ok. I really dont know much about video and audio formats etc. [14:56:18 CEST] <furq> obviously if you were using transpose you'd be reencoding, in which case it'd be faster to resize [14:56:24 CEST] <furq> but this is just changing some metadata in the container [14:57:14 CEST] <furq> it doesn't actually rotate the video, it just instructs the player to do so [15:01:20 CEST] <n1cksn1ck> furq: please forgive my silence/slow response to you. Because i really dont know much about all this video/audio codec/format etc. Im finding it hard how to word my questions or responses to your comment. do you know what i need to type to transpose it 90deg? [15:01:34 CEST] <adamk___> i found a video which i thought looked really good, is it possible to find which settings and filters the encoder used? [15:02:01 CEST] <furq> adamk___: mediainfo may show you depending on which encoder was used [15:02:29 CEST] <n1cksn1ck> furq: Im up to 3800 frames. just over 2000 frame to go : ) [15:02:50 CEST] <furq> n1cksn1ck: if you're using transpose it should work fine [15:03:23 CEST] <furq> -metadata:s:v:0 rotate=90 may or may not work but it should be more or less instant [15:04:00 CEST] <adamk___> furq: it was encoded in x264, i see there are encoding settings. is there a way i can use the same ones? [15:04:09 CEST] <n1cksn1ck> im not using it(yet). I dont know how : ( can you please give me an example like you did with rotate? [15:04:28 CEST] <furq> n1cksn1ck: pastebin the command you're using [15:04:33 CEST] <furq> adamk___: copy and paste [15:04:58 CEST] <furq> or pastebin them and i'll tell you if it's using one of the standard presets, which it probably is [15:06:34 CEST] <adamk___> furq: http://pastebin.com/VuUiC6mg [15:07:41 CEST] <furq> well it's not a standard preset [15:08:15 CEST] <furq> but it's also using 2-pass which means the guy who encoded it probably isn't to be trusted [15:08:31 CEST] <furq> either that or he distributed this movie in a bunch of 100MB rar files [15:09:05 CEST] <n1cksn1ck> furq: I dont know pastebin. but when i tried to use the transpose command. I just added transpose=1. [15:09:12 CEST] <adamk___> why is he not to be trusted? [15:09:32 CEST] <furq> 2-pass is pretty much useless unless you're targeting a specific filesize, which is basically worthless because it's 2016 [15:09:56 CEST] <furq> the only reason to do that is if you don't want your perfectly legal release to be nuked off topsites [15:10:23 CEST] <furq> fwiw i just encode everything with -preset slow -crf 20 [15:10:23 CEST] <adamk___> i see [15:10:41 CEST] <furq> there's not much to be gained from tweaking settings like that [15:10:59 CEST] <furq> everything will look great with a low enough crf [15:12:13 CEST] <furq> n1cksn1ck: -vf transpose=1 [15:12:24 CEST] <furq> that should rotate 90 degrees clockwise [15:13:33 CEST] <adamk___> furq: i have the bluray source of the same file http://imgur.com/a/3OIuY [15:14:04 CEST] <adamk___> i have no idea how he did this [15:14:27 CEST] <furq> is the source the bottom one [15:14:34 CEST] <adamk___> yes [15:14:39 CEST] <n1cksn1ck> furq: i think i did try that also(found it at a puppylinux forum) got an error. so tried what i could from ffmpeg Documentation. Also got an error. maybe i typed something wrong. This file is nearly finished : ) [15:14:57 CEST] <furq> i don't think that has anything to do with x264 settings then [15:15:28 CEST] <adamk___> then what did he do [15:15:35 CEST] <furq> probably filters [15:15:43 CEST] <furq> certainly for the colour balance [15:15:56 CEST] <furq> or maybe he just has a better source [15:15:57 CEST] <adamk___> and i can't see which ones he used? [15:16:09 CEST] <furq> not without asking him [15:16:22 CEST] <n1cksn1ck> furq: the ouput video wasnt rotated : ( [15:16:22 CEST] <adamk___> i see [15:16:42 CEST] <adamk___> but i can apply filters through ffmpeg, correct? [15:17:10 CEST] <n1cksn1ck> furq: i will try transpose(how you've suggested). [15:18:03 CEST] <furq> adamk___: https://ffmpeg.org/ffmpeg-filters.html [15:20:12 CEST] <adamk___> furq: thanks, i will try to dabble a little bit in that [15:25:46 CEST] <n1cksn1ck> furq: tried -vf transpose=1 and got an unrecognized option 'vf' in the generated log file. Any suggestions? [15:26:21 CEST] <furq> pastebin the command [15:27:38 CEST] <n1cksn1ck> furq: i dont know how. [15:29:02 CEST] <furq> http://pastebin.com/ [15:31:13 CEST] <n1cksn1ck> furq: do you mean all the stuff in the Adv options Box? [15:31:47 CEST] <n1cksn1ck> and does it matter if im using ffconvert v1.2? [15:38:06 CEST] <n1cksn1ck> hey furq: im need a lil break. thanks heaps for your time and suggestions. I'm going to make some food. Grab a cuppa ; ) If i dont talk with you again. I hope you have a great day. [15:38:13 CEST] <ibrasiskis> Hello does anyone knows if the hap codec in ffmpeg is hardware accelerated? [15:42:22 CEST] <ibrasiskis> That is uses opengl to decompress frame textures? [15:47:02 CEST] <jkqxz> ibrasiskis: No, it's purely software. [16:03:18 CEST] <ibrasiskis> It sucks a little, because that is the point of the codec. It would be adventure to port the codec to linux myself, but probably that is over my capabilities as OpenGL/multimedia programmer. :) [16:07:53 CEST] <IanWizard> ibrasiskis: what codec? [16:11:26 CEST] <ibrasiskis> Hap codec, I asked if it uses gpu to decompress frames on ffmpeg, but I was told that it is purely software based. [16:17:44 CEST] <ibrasiskis> Hap is Intraframe (keyframe only) codec specialiazed for visualisations and projections, encodes alpha channel and is normally used to compose scenes from multiple scenes. [17:01:19 CEST] <yelmond> hello - is there a way to convert image files (say, .nef raws) on-the-fly using another program as ffmpeg processes them into a video? Trying to avoid having to convert ALL the images first and then running ffmpeg. [17:03:26 CEST] <c_14> Assuming the program to convert the images can do so in sequence and output them on stdout, yes [17:03:32 CEST] <c_14> use ffmpeg's image2pipe demuxer [17:05:37 CEST] <yelmond> thanks c_14! I'll look into image2pipe. [17:22:52 CEST] <mundus2018> Can someone help me convert a bunch of .m2ts files to a lower bitrate/quality [17:23:17 CEST] <mundus2018> right now like an hour and a half video is about 6GB [17:23:36 CEST] <mundus2018> Im shooting for 700mb-1gb [17:27:14 CEST] <n1cksn1ck> thanks again furq: [17:30:47 CEST] <pgorley> Is ffmpeg_videotoolbox.c an example on how to use vt or code I need to include to use it? [17:32:59 CEST] <c_14> mundus2018: https://trac.ffmpeg.org/wiki/Encode/H.264 [17:33:22 CEST] <c_14> pgorley: pretty sure it's an example [17:33:29 CEST] <jkqxz> Mainly an example. The ffmpeg_*.c hwaccel files are implementation-specific bits needed to use that hwaccel in the ffmpeg utility (combined with the other bits in ffmpeg.c). [17:33:52 CEST] <pgorley> Alright, thanks! [18:06:50 CEST] <Sashmo> can anyone tell me whats the best way to overlay a animated logo on top of a video? or better yet an swf or flv with alpha on top of the video, ideally looking at adding dynamic data on top of my live stream [18:08:27 CEST] <Angus> Trying to create a custom video decoding plugin here. What does hwaccel_device do? [18:30:00 CEST] <neuro_sys> Is there a magick trick to speed up ffmpeg (increasing threads perhaps)? [18:30:22 CEST] <neuro_sys> I have 5 input streams one complex filter with 5 to 10 chains and one output stream [18:43:22 CEST] <BtbN> Add more cores, reduce quality. [18:43:22 CEST] <nMaib0> how can I convert an audio file from 24 fps to 23.976? [18:43:32 CEST] <nMaib0> dts to opus [18:43:42 CEST] <BtbN> Audio does not have video-fps. [18:44:45 CEST] <nMaib0> yeah tell that to the video I just converted from 24 to 23.976 and now the audio is out of sync, I need to speed it up to match the resultant video [18:45:10 CEST] <nMaib0> I could fix it by telling matroska to play the video at 24 but that's a crappy fix [18:45:16 CEST] <BtbN> so you didn't convert the video, you just speeded it up. [18:46:34 CEST] <nMaib0> no I converted it to 23.976 [18:47:12 CEST] <nMaib0> maybe I converted it wrong... [18:47:53 CEST] <Mavrik> If your audio is out of sync then you just sped up the video without actually doing the pulldown :/ [18:47:54 CEST] <nMaib0> I read that most BD 24fps files are in reality 23.976 so I used assumefps on avisynth, like some user advicced on doom9 [18:48:39 CEST] <Mavrik> Well it seems it wasn't :P [18:48:43 CEST] <nMaib0> yeah, but I find that to be a crappy fix, matroska can do that easily and I already tested and it works. [18:48:52 CEST] <nMaib0> yeah haha [18:49:26 CEST] <Mavrik> So either you use atempo to speed up audio a bit [18:49:43 CEST] <Mavrik> Or do the conversion with actually changing fps [18:50:38 CEST] <nMaib0> yeah I think I'll leave the 24fps [18:50:46 CEST] <nMaib0> and reconvert again [18:53:24 CEST] <casf> i'm streaming AVFrames through an AVFilterGraph and would to change filter parameters mid-stream. is there a recommended way to do this or an example somewhere? [18:56:53 CEST] <durandal_1707> casf: only by recreating graph [18:57:48 CEST] <casf> hrmmm [18:58:08 CEST] <casf> i'm worried about gaps, but i haven't even tried it yet [18:58:17 CEST] <casf> or like...dropped frames from the buffer sink [18:58:42 CEST] <casf> i guess i could just reuse the buffer sink? freeing the filter graph just destroys the links and leaves the filter context objects unmodified. [19:02:46 CEST] <casf> durandal_1707: have you implemented anything similar that rebuilds a filter graph mid-stream? [19:07:01 CEST] <BtbN> some filter support runtime reconfig [19:07:23 CEST] <casf> BtbN: are there any docs for that? [19:07:31 CEST] <BtbN> I'd guess so. [19:11:28 CEST] <vade> hello. Im trying to flush my resample context via swr_convert_frame . the doc header state NULL input is how to do this, but I get a crash [19:11:31 CEST] <vade> is this expected? [19:12:04 CEST] <vade> btw Mavrik fixed my issue the other day - was not setting duration on packets to the muxer so last frame had 0 duration, thats why frame counts matched byt duration / fps didnt. [19:21:12 CEST] <durandal_1707> casf: I had [19:21:30 CEST] <casf> durandal_1707: found this https://github.com/PromyLOPh/pianobar/blob/master/src/player.c -- good example of dynamic volume [19:21:45 CEST] <casf> that coupled with BtbN's recommendation that _some_ filters support runtime reconfig is super helpful [19:21:48 CEST] <casf> thanks to you both. :D [19:22:45 CEST] <casf> i'm guessing i'll eventually need to deal with the graph reconstruction, but for now this is a good starting point for me - just getting my feet wet with ffmpeg libs anyway [19:23:28 CEST] <casf> and sorry, lines 81-133 have the example since that is a pretty big file [19:26:20 CEST] <vade> Yea, im not sure how im supposed to vend samples from my resample context via swr_convert_frame - i cant provide a NULL input without crashing, and if I make an empty frame with appropriate format, channel and sample rate, i keep spewing out empty packets [19:26:29 CEST] <vade> rather, empty frames. [19:29:35 CEST] <tobor> So I am sitting here at work on a windows_10 computer and know nada bout this os. See there is a static build for ffmpeg on windows. Need to import video from a DV video camera to a via firewire. Want to drop this directly to our 18TB ZFS NAS. Can this be done using a windows_10 os? [19:32:49 CEST] <tobor> my output ding ffmpeg -f dshow -list_devices true -i dummy [19:32:52 CEST] <tobor> http://paste.debian.net/683428/ [19:37:39 CEST] <tobor> wow gonna be nice in a few weeks when GNU/Linux will be incorporated into windows_10. The idea of doing something like compiling with gcc in a windows shell or something like - watch -n1 'grep Mhz /proc/cpuinfo |sort -k4rn' in windows seems awesome. [19:39:09 CEST] <tobor> Not sure how that is working out with the GPL though. Seems Can Canonical worked something out with MS. [19:40:28 CEST] <tobor> lol can just apt-get directly in a windows bash shell and use Canonical's (ubuntu's) repos. [19:41:27 CEST] <kiroma> Hello [19:41:36 CEST] <tobor> lol can just apt-get directly in a windows bash shell and use Canonical's (ubuntu's) repos in a few weeks. [19:42:00 CEST] <tobor> sup kiroma [19:42:34 CEST] <tobor> and yes I did say bash. [19:42:39 CEST] <kiroma> I am a beggining linux user, where should I put header files to configure ffmpeg properly? [19:44:01 CEST] <tobor> kiroma, I wrote this a few years back. http://pastebin.com/NHXrsBxm [19:44:15 CEST] <tobor> prolly does not work today [19:44:24 CEST] <tobor> what distro you on? [19:45:08 CEST] <kiroma> linux mint 17.3 and ffmpeg 3.0.2 [19:45:20 CEST] <kiroma> I wanted to configure it so that I could use nvenc [19:47:11 CEST] <tobor> in the past I would use checkinstall --pkgname=x264 --pkgversion="3:$(./version.sh | awk -F'[" ]' '/POINT/{print $4"+git"$5}')" --backup=no --deldoc=yes --fstrans=no --default [19:47:42 CEST] <ChocolateArmpits> tobor: wait does that mean there will be a way to build ffmpeg without using mingw/msys ? [19:47:58 CEST] <tobor> yeah! [19:48:03 CEST] <tobor> freaking awesome! [19:48:04 CEST] <ChocolateArmpits> oh snap [19:48:22 CEST] <ChocolateArmpits> But it's W10 only ? [19:48:32 CEST] <tobor> for poor a fsck like me that cant use GNU/Linux at work [19:49:13 CEST] <tobor> just type in " bash " in shell and poof you have a REAL linux shell. [19:49:47 CEST] <c_14> kiroma: --extra-cflags='-I/path/to/headers' [19:50:08 CEST] <c_14> Or just throw the header in /usr/local/include (that's usually in the default include path) [19:50:20 CEST] <kiroma> Oh, allright. Thanks! [19:51:19 CEST] <tobor> lol take a look at this. GNU/Linux by default in windows_10 https://www.youtube.com/watch?v=nKtNFZRWO0U [19:51:40 CEST] <tobor> using gcc or aptitude and what now. Nuts [19:53:54 CEST] <tobor> The DD's over at Debian to MS to f**k off but Canonical grabbed their toes and told MS to let her rip! [19:54:51 CEST] <tobor> ChocolateArmpits, not sure if it is only w10 [19:55:40 CEST] <kepstin> it is only win 10, afaik. It's basically like wine except the other way around. They added a compat layer that lets you run native linux apps on the windows nt kernel. [19:56:36 CEST] <kepstin> and they ship it with an (apparently unmodified) ubuntu userspace, but in theory you can run anything with it. [19:56:41 CEST] <tobor> but how the heck and say " watch -n1 'grep MHz /proc/cpuinfo | sort -k4rn' work? I mean /proc?????? [19:57:45 CEST] <kepstin> '/proc' is part of the linux kernel api, they're probably simulating it for app compatibility reasons. [19:57:46 CEST] <tobor> or setting up IPT tables and making a real firewall. That is kernel stuff [19:58:05 CEST] <kepstin> it almost certainly doesn't support iptables. [19:58:17 CEST] <tobor> from what I hear it will???? How [19:58:24 CEST] <kepstin> (you can of course run a full virtual machine with a linux kernel if you want to do that) [19:58:42 CEST] <tobor> blah only use MS stuff at work. [20:00:13 CEST] <tobor> hmmm wonder what say uname -a would show? [20:03:04 CEST] <tobor> what ever. Right now I am just mad that does not exist yet on this windows_10 machine. I don't know windows and I am trying to import video via firewire to a ZFS nas using ffmpeg [20:03:21 CEST] <tobor> dshow? [20:04:29 CEST] <tobor> kepstin can I pass though this firewire card with something other them kvm? I use a lot of vfio and qemu at home [20:11:30 CEST] <durandal_1707> casf: dynaudnorm filter? [20:15:34 CEST] <tobor> right this is not working - ffmpeg -f iec61883 -i auto -hdvbuffer 100000 testout.mpg . right so I take it libiec61883 is not how it is done in windows. [20:16:43 CEST] <tobor> what in heck is used here? Not the stack juju [20:18:12 CEST] <casf> durandal_1707: not yet - just tinkering with volume right now [21:26:19 CEST] <neuro_sys> I have this command generated programmatically, but it takes 50 seconds. Any chance I could reduce it down? [21:26:22 CEST] <neuro_sys> https://gist.github.com/neuro-sys/1d857664c5a32cabae4a2b1d0a826a60 [21:27:20 CEST] <neuro_sys> the parts of the video where there's no overlaying (filtering) involved maybe is causing useless demux/remux/decode/encode phases. I'm not sure how it really works internally in that regard. [21:29:51 CEST] <ChocolateArmpits> neuro_sys: if that's your command then there is encoding happening after the filter [21:31:03 CEST] <neuro_sys> ChocolateArmpits: Yes it's the command. [21:31:11 CEST] <neuro_sys> could you elaborate? [21:31:54 CEST] <ChocolateArmpits> if you don't specify encoding options ffmpeg will then use defaults. Depending on the video format that can result in slow encoding process [21:31:59 CEST] <neuro_sys> what I'm trying to achieve is to overlay 5 alpha image streams on top of the first video stream [21:32:20 CEST] <ChocolateArmpits> Your command isn't just overlaying the video but also encoding the result [21:32:24 CEST] <neuro_sys> but overlays happen only in certain parts of the video; hence enable=between(n,a,b) [21:32:44 CEST] <neuro_sys> isn't encoding the result naturally required? [21:33:20 CEST] <ChocolateArmpits> well yeah but you can pick different formats or speed up the encoding by setting proper parameters [21:33:56 CEST] <ChocolateArmpits> as it is it's encoding to h264 and probably using medium preset [21:34:07 CEST] <ChocolateArmpits> if the video is hd medium will be quite slow [21:34:10 CEST] <neuro_sys> It's supposed to be viewed in a browser, and I'm not sure which would be the optimal then [21:34:45 CEST] <ChocolateArmpits> what's the resolution and framerate of the output ? [21:35:06 CEST] <neuro_sys> I was thinking if I should just overlay each part of the video seperately using -ss and -t, and then concatenate them to avoid useless decoding/encoding or whatnot if that makes any sense [21:35:29 CEST] <neuro_sys> because in 60 seconds of video (25 fps: 640x350), there're only 15-20 seconds of overlaying parts [21:35:41 CEST] <ChocolateArmpits> that's impossible [21:35:56 CEST] <neuro_sys> I understand [21:36:09 CEST] <neuro_sys> so it would not have any speed gain, or just not possible? [21:36:18 CEST] <ChocolateArmpits> of course it's possible [21:37:36 CEST] <ChocolateArmpits> anyways try this after the map command and before the output "-vcodec libx264 -g 50 -preset veryfast -b:v 900k -vprofile main -movflags faststart -chunk_duration 1000k" [21:37:56 CEST] <neuro_sys> okay, I'm trying now [21:38:26 CEST] <ChocolateArmpits> faststart will move the index to the beginning of the file, chunk_duration I think set the chunks of the index in sizes of 1 second here [21:42:09 CEST] <neuro_sys> wow! [21:42:22 CEST] <ChocolateArmpits> any faster ? [21:42:34 CEST] <neuro_sys> yeah, it was in 25 seconds now! [21:42:39 CEST] <neuro_sys> rather than 50 seconds [21:42:56 CEST] <neuro_sys> I have no idea how that works, but I'll study them [21:43:11 CEST] <ChocolateArmpits> do you have specific bitrate requirements ? [21:43:40 CEST] <neuro_sys> no I'm not told any requirement [21:43:47 CEST] <c0rnw19> hello world [21:44:08 CEST] <neuro_sys> I think 1mbp/s but I'll check [21:44:44 CEST] <ChocolateArmpits> well you can try -vcodec libx264 -g 25 -preset superfast -b:v 1000k -vprofile baseline [21:44:48 CEST] <ChocolateArmpits> this will be faster [21:44:58 CEST] <ChocolateArmpits> but the quality will end up worse [21:45:01 CEST] <ChocolateArmpits> than the previouis [21:45:14 CEST] <ChocolateArmpits> try for yourself [21:45:16 CEST] <furq> does -g really make a difference to encoding speed [21:45:17 CEST] <neuro_sys> hmm, which part reduces the quality in this one? the previous command had -b:v 900k, and this has higher [21:45:32 CEST] <ChocolateArmpits> furq: well it has to look up less frames [21:45:41 CEST] <neuro_sys> and what aobut -chucnk_duration? [21:46:28 CEST] <ChocolateArmpits> neuro_sys: you can add movflags and chunk_duration but your video isn't long so those can be omitted, however they are very important for long videos [21:48:18 CEST] <ChocolateArmpits> neuro_sys: the specific commands for performance/quality control are "-g" controls group of picture size, "preset" sets the quality versus speed of the encoder, "-b:v" sets video bitrate, "vprofile" sets the target profile for the encoder, simpler profiles utilize fewer tricks so they are faster [21:49:02 CEST] <ChocolateArmpits> the important profiles h264 has are baseline, main and high [21:49:35 CEST] <ChocolateArmpits> you can see the preset explanation here http://dev.beandog.org/x264_preset_reference.html [21:49:35 CEST] <furq> if you use baseline then you might as well just use preset ultrafast [21:49:54 CEST] <ChocolateArmpits> well ultrafast should not advised under most circumstances [21:50:02 CEST] <ChocolateArmpits> it turns off way too much stuff [21:50:04 CEST] <furq> neither should baseline [21:50:07 CEST] <ChocolateArmpits> look at the table [21:50:28 CEST] <ChocolateArmpits> hey mobiles need that baseline [21:50:43 CEST] <ChocolateArmpits> and if you're running on a toaster you can make use of it too [21:51:11 CEST] <furq> they both turn off cabac and partitions, which are a big deal [21:51:46 CEST] <ChocolateArmpits> the guy needs speed, what can I else suggest [21:52:08 CEST] <ChocolateArmpits> suggesting ultrafast I won't however [21:52:49 CEST] <furq> i wouldn't suggest baseline either unless you need this video to run on an iphone from 2007 [21:53:24 CEST] <furq> neuro_sys: changing the preset will make the biggest difference by far [21:54:15 CEST] <furq> you also want to use -crf instead of -b:v [21:54:37 CEST] <furq> if you have specific bitrate needs then either use the vbv (bad quality) or use 2-pass (slow) [22:06:05 CEST] <kbarry> I'm trying to better understand some output in ffmplay -loglevel 56: ch:2 chl:stereo fmt:fltp r:44100Hz -> ch:1 chl:mono fmt:s16 r:44100Hz [22:09:14 CEST] <neuro_sys> as of which version of ffmpeg there's -antialias option available? [22:09:41 CEST] <kbarry> I'd like to better understand the fmt: portion of the output, [22:10:13 CEST] <kbarry> I cannot find a lot of information, and might just lack the right words to google. Anyone mind pointing me in the right direction to search? [22:15:37 CEST] <casf> kbarry: https://ffmpeg.org/ffplay.html#toc-Main-options -f is under there [22:15:59 CEST] <kbarry> casf: Thanks, let me go do some reading [22:16:25 CEST] <vade> under what circumstances would the start time of a stream as reported by ffprobe not be zero ? [22:18:16 CEST] <ChocolateArmpits> vade: from my experiences, timestamp information can be taken from a clock so the start time will be the start time since the operation has started [22:18:31 CEST] <ChocolateArmpits> also it can that an audio stream starts earlier than a video stream so the start time will be negative [22:19:13 CEST] <ChocolateArmpits> my experience [22:19:59 CEST] <vade> yea, start isnt negative, its a super small value, 0.021333 - and my video and audio packets appear to have PTS and DTS iniciating zero relative to the muxer when I write them [22:20:03 CEST] <casf> kbarry: this might also help, but it's a lot more reading: https://ffmpeg.org/ffmpeg-formats.html [22:20:30 CEST] <vade> im also setting start_time to zero for my streams [22:22:27 CEST] <kbarry> casf: Thanks. There are two terms I am running into, "fltp" and "s16", I think i understand these for bemore or less, how the bits are arranges, but I can't find any informations that might shed light on the "why you would use one over the other, or what the affects of using s16 over X " [22:22:50 CEST] <kbarry> I'm only dealing with audio, and have also found that to be a roadblock. [22:29:39 CEST] <durandal_1707> kbarry: that depends on how codec works [22:30:10 CEST] <kbarry> Mind expounding a little? [22:30:37 CEST] <kbarry> you mean things like , X codec supports s16, but not fltp ? [22:30:46 CEST] <furq> most audio codecs only support one or the other [22:32:55 CEST] <furq> it's unlikely to be something you'll ever have to consider [22:34:57 CEST] <vade> oh interesting [22:35:04 CEST] <vade> looks like ffprobe clips the negagive [22:35:27 CEST] <vade> if I do ffprobe -show_strams -select_streams a -infile- [22:35:33 CEST] <vade> I get start_time - 0.021333 [22:35:45 CEST] <kbarry> fruq, I am only asking about it because I am "somehow" producing a stream where the format is changing mid-stream (the stream is being produced ouside of ffmpeg), the stream is going from mono to stero, from stero to mono, from fltp to s16, etc. [22:35:51 CEST] <vade> but pure ffpobe shows me positive start time 0.02133 [22:36:24 CEST] <furq> afaik if ffprobe shows fltp -> s16 then that's just a result of it decoding a lossy format [22:36:38 CEST] <furq> i don't know why it would be changing channel layout, though [22:36:57 CEST] <furq> s/ffprobe/ffplay/ [22:37:26 CEST] <kbarry> The short version is "Some other tool;" is taking file.mp3, and foo.mp3, and urlstream, and broadcasting them as an RTMP stream, one after anoter, [22:37:32 CEST] <kbarry> lets say each changes every 30 seconds. [22:38:22 CEST] <kbarry> file.mps is mono, fltp, foo.mp3 is stero, s16. as is the url. [22:39:07 CEST] <kbarry> When a player (like ffplay) tunes into the rtmp stream, everything is cool, until the content, on the back end, changes, and the specs of the audio change, [22:39:22 CEST] <kbarry> it started mono, fltp, and not the stream is stero s16 [22:39:30 CEST] <kbarry> not = now [22:47:16 CEST] <neuro_sys> is there a rough performance comparison chart for video encoding formats [22:51:18 CEST] <furq> not that i know of, but you won't get much faster than x264 [22:52:22 CEST] <darsie> is x264 better than xvid? [22:52:28 CEST] <furq> yes [22:52:31 CEST] <darsie> thx [23:07:28 CEST] <vade> so heres a question. Im using libswresample to resample audio typically from 44.1 to 48 khz - my limited understanding is that libswresamples swr_convert_frame will keep an internal queue for samples since - well, it has to [23:08:01 CEST] <vade> however, im getting more samples than frame size from my encoder - [23:09:27 CEST] <vade> my swr_convert_frame expectedly works if im going 48 -> 48 Khz - however, actually trying to call swr_convert_frame doesnt throw and error and I get a valid frame out from what I can tell [23:09:57 CEST] <vade> do I need to use an AudioFrameQueue post resample just due to frame size nuances? [23:19:35 CEST] <vade> yea, so my input frame pre resamples has a frame size of 1024 [23:20:19 CEST] <vade> i resample, its is 1098 [23:20:32 CEST] <vade> my codec context is frame_size is 1024 [00:00:00 CEST] --- Thu May 12 2016

1 0