Ffmpeg-devel-irc

Download

ffmpeg-devel-irc@ffmpeg.org

August 2016

1 participants
62 discussions

ffmpeg-devel.log.20160731
by burek 01 Aug '16

01 Aug '16

[04:52:52 CEST] <cone-460> ffmpeg 03Timothy Gu 07master:5522da9ee102: doxygen: Remove outdated EXPAND_AS_DEFINED [04:52:52 CEST] <cone-460> ffmpeg 03Timothy Gu 07master:cebad64d4f06: Doxyfile: Reset unused options back to default [07:51:58 CEST] <KGB> [13FFV1] 15TimothyGu pushed 1 new commit to 06master: 02https://git.io/v6eqX [07:51:58 CEST] <KGB> 13FFV1/06master 14da40880 15Reto Kromer: Fix dead link of HuffYUV (#23) [11:11:41 CEST] <kierank> atomnuker: ^ [16:39:47 CEST] <cone-091> ffmpeg 03Burt P 07master:999c6b1da85a: af_hdcd: fix a minor cosmetic annoyance [16:39:47 CEST] <cone-091> ffmpeg 03Burt P 07master:390eea61be59: af_hdcd: give cdt expired counter a value for never set [16:54:54 CEST] <bms20> Hello, I'm trying to determine if a video is interlaced or not. I'm looking at the field_order entry in AVCodecParserContext and AVCodecContext - both are AV_FIELD_UNKNOWN on two test videos I've tried. Is there something I'm doing wrong, or is this information just not coded in the files? [16:55:33 CEST] <nevcairiel> that information is typically not available on a per-file basis [16:57:41 CEST] <bms20> Is the information available on a per-frame basis? [16:57:47 CEST] <bms20> Or is it just not stored? [16:57:52 CEST] <bms20> (in any file)? [16:58:01 CEST] <nevcairiel> per-frame yes [16:58:11 CEST] <nevcairiel> AVFrame has it [16:59:12 CEST] <bms20> Ok - cool. Am looking now. [17:00:08 CEST] <bms20> ok - think I've got it. interlaced_frame and top_field_first. [17:00:11 CEST] <bms20> Thanks for that! [17:38:13 CEST] <bp0> michaelni, Alright, what do I need to do to assume 'maintainer of hdcd'? [17:39:44 CEST] <michaelni> bp0, send a patch that adds you to the MAINTAINERs file for hdcd, if noone objects and the patch is applied then you are maintainer of it [17:41:30 CEST] <bp0> ok [17:48:19 CEST] <bp0> done. [18:34:31 CEST] <cone-091> ffmpeg 03Michael Bradshaw 07master:c9ae8be5a87c: avformat/matroskaenc: fix Voids with size < 10 [18:35:28 CEST] <Timothy_Gu> michaelni: can you test this patch: http://sprunge.us/BUMU? I can't reproduce that bug here but the GCC documentation does show a few errors in my original patch [18:41:51 CEST] <Bitterblue> Hi. I see that the PGS subtitle decoder doesn't return an AVSubtitle for every packet fed to avcodec_decode_subtitle2, even though the decoder doesn't have CODEC_CAP_DELAY. What is going on there? [18:42:17 CEST] <nevcairiel> those things are not necessarily related [18:42:33 CEST] <nevcairiel> sometimes you just need to assemble more then one packet to be able to produce output [18:43:24 CEST] <nevcairiel> delay just means it needs to be flushed at the end, which might not apply as partial data may not be decodeable at all stand-lone [18:43:27 CEST] <nevcairiel> +a [18:43:57 CEST] <Bitterblue> I see. [18:47:00 CEST] <michaelni> Timothy_Gu, make: *** [fate-lossless-shorten] Error 134 [18:52:13 CEST] <michaelni> seems not reproduceable under gdb with debug symbols [19:00:05 CEST] <michaelni> Timothy_Gu, backtrace: http://pastebin.com/vccsQqFw [19:02:26 CEST] <Timothy_Gu> michaelni: thanks, I'll just drop this patch. I can't repro on gcc (Debian 4.9.2-10) 4.9.2 and it's not worth the effort [19:19:22 CEST] <cone-091> ffmpeg 03Timothy Gu 07master:04da20e58fd7: dynarray: Change AV_ to FF_ for AV_DYNARRAY_ADD [19:21:37 CEST] <omerjerk> hey is this function supposed to be executed when encoding from wav to mp4 als? - https://github.com/omerjerk/FFmpeg/blob/alsenc/libavformat/alsenc.c#L83 [19:21:49 CEST] <omerjerk> the log line isn't coming though. [19:22:03 CEST] <omerjerk> The encoder gives seg fault as well. :| [19:22:28 CEST] <omerjerk> it completes the encoding though. I receive the last frame and then after that somewhere is segfaults. [19:22:36 CEST] <omerjerk> Any tips on how can I debug this? [20:06:55 CEST] <Bitterblue> Huh, every other subtitle I get from the PGS decoder has num_rects=0. [20:20:11 CEST] <nevcairiel> its probably to indicate that the display should be cleared [20:21:50 CEST] <Bitterblue> Yeah, the packet pts does indicate that. [20:22:47 CEST] <Bitterblue> Which is weird, because AVSubtitle has those start_display_time and end_display_time fields. [20:23:06 CEST] <Bitterblue> Set to useless values, in this case. [20:36:09 CEST] <Bitterblue> But if several packets are needed to produce a subtitle, and they all have different pts, which of those should be used as start time? [20:38:53 CEST] <michaelni> Does anyone want FFmpeg to participate in the next Outreachy ? [20:39:52 CEST] <michaelni> if so theres need to create an ideas page: https://trac.ffmpeg.org/wiki/SponsoringPrograms/Outreachy/2016-12 [20:40:48 CEST] <michaelni> similar to previous ones: https://trac.ffmpeg.org/wiki/SponsoringPrograms/Outreachy/2016-05 or https://trac.ffmpeg.org/wiki/SponsoringPrograms/Outreachy/2015-05 or https://trac.ffmpeg.org/wiki/SponsoringPrograms/GSoC/2016 [20:49:17 CEST] <bp0> patchwork.ffmpeg.org seems very good [20:56:26 CEST] <kierank> michaelni: can we actually come up with new takmsks instead of the same crap each year [20:58:05 CEST] <michaelni> kierank, new tasks would be ideal [20:58:30 CEST] <kierank> Yes but don't put the same old junk in [20:58:50 CEST] <kierank> Because it's so boring [21:00:34 CEST] <michaelni> we need mentor task pairs, if we have mentors that want to mentor new tasks i would favor that, if we have nothing but old ideas and mentors who are happy mentoring them then thats the only thing we can put on the page [21:05:12 CEST] <nevcairiel> if we only have those tasks that have lived on that page for 5 years and noone ever wanted to do them, maybe we should just not join [21:08:57 CEST] <michaelni> what would you suggest for fresh tasks ? [21:14:28 CEST] <jamrial> BBB mentioned some weeks ago an idea he had for outreachy i think [21:14:36 CEST] <jamrial> but in general most ideas people had were considered too complex for this kind of thing [21:14:46 CEST] <BBB> I forgot what it was already [21:15:02 CEST] <jamrial> because they included designing a new API and such [21:15:04 CEST] <BBB> but I agree that we should get rid of old tasks after N years, where N is small [21:15:20 CEST] <BBB> if a list of tasks just contains stuff nobody ever wanted to do, its likely nobody actually wants to do it [21:15:25 CEST] <BBB> and therefore it just pushes people away [21:15:31 CEST] <BBB> theyll look for other, more appealing, task lists [21:17:22 CEST] <BBB> was what I wanted to do the vp9 loopfilter? I dont think it was [21:17:28 CEST] <BBB> does anyone have a log of what I proposed? [21:17:53 CEST] <jamrial> BBB: i'm trying to find it [21:18:16 CEST] <jamrial> should be somewhere in the irc logs [21:19:34 CEST] <BBB> I hope so :-p [21:21:19 CEST] <jamrial> https://ffmpeg.org/pipermail/ffmpeg-devel-irc/2016-July/003682.html [21:21:30 CEST] <jamrial> it was vp8/9 alpha support on the native decoders [21:42:36 CEST] <iive> well, that's 1 and it is from this month. Do you have any other good ideas? [00:00:00 CEST] --- Mon Aug 1 2016

1 0

ffmpeg.log.20160731
by burek 01 Aug '16

01 Aug '16

[03:02:46 CEST] <medya> can you concatanate multiple .avi files and at the asme time convert outpfile to mp4 [03:03:00 CEST] <medya> so lets say you have 3 avi files, you want one mp4 file [03:03:11 CEST] <medya> how to combine concatante and convert to mp4 at same time? [03:19:35 CEST] <DHE> there's a concatenate input format that will assemble the inputs. see the trac wiki [03:23:09 CEST] <medya> what is trac wiki [03:24:17 CEST] <medya> I dont see one here https://trac.ffmpeg.org/wiki/Concatenate [03:24:20 CEST] <medya> it is all -copy [03:24:23 CEST] <medya> I dont want copy [03:25:08 CEST] <DHE> then scroll down to the "different codecs" bit [03:25:18 CEST] <DHE> also note that's codec copy, not container copy [03:27:34 CEST] <medya> DHE thats for different codes of input [03:27:40 CEST] <medya> all my inputs are avi [03:27:46 CEST] <medya> I just want diffrent ouptput [03:27:57 CEST] <medya> the doc talks about if I have bunch of files with diffrent codecs [03:32:58 CEST] <DHE> I fear you are confusing "codec" and "container". h264, mpeg2 and sorenson are codecs. avi, mp4 and webm are containers. [03:39:03 CEST] <medya> well I have 3 files, with avi contrainers, I want all to be concatanted into one file with h264 codec and mp4 container [03:39:06 CEST] <medya> how can I do that? [04:18:30 CEST] <buff3r> hi trying to satisfy the output format for a pipe [04:18:41 CEST] <buff3r> (trying to save a png clip from a video) [04:19:24 CEST] <buff3r> does anyone know how to specify a png for pipe:1? [04:19:25 CEST] <bp0> satisfy = specify? [04:19:40 CEST] <buff3r> i tried -f png_pipe but that didn't work [04:19:44 CEST] <buff3r> yes [04:19:50 CEST] <furq> -f image2pipe [04:20:13 CEST] <furq> it doesn't sound like you need a pipe though [04:20:34 CEST] <buff3r> i think that worked [04:21:14 CEST] <buff3r> furq: yeah but i plan on using this as part of a python program [04:21:36 CEST] <buff3r> bp0: yes specify, thanks [04:40:28 CEST] <zumba_addict> hi folks, i ran ffmpeg -i on an mp4 and I got this - Stream #0:0(eng): Video: hevc (Main) (hvc1 / 0x31637668), yuv420p(tv), 1280x720 [SAR 1:1 DAR 16:9], 139 kb/s, 29.97 fps, 29.97 tbr, 30k tbn, 29.97 tbc (default) [04:40:38 CEST] <zumba_addict> I can't see the video but I can hear the audio loud and clear [04:40:55 CEST] <zumba_addict> what is that format? I've never seen one like that [04:43:14 CEST] <furq> h.265 [04:45:51 CEST] <zumba_addict> looks like i need to update mplayer to play it [04:45:58 CEST] <zumba_addict> vlc was able to play it [04:46:16 CEST] <zumba_addict> is it better than h.264? [04:46:31 CEST] <bp0> in ways [04:46:50 CEST] <zumba_addict> k [04:55:56 CEST] <zumba_addict> WOW, https://www.youtube.com/watch?v=qL22L0mRSDs [04:56:03 CEST] <zumba_addict> if it's true, WOW [04:56:39 CEST] <zumba_addict> how can I check if my ffmpeg has support for hevc? [04:57:01 CEST] <zumba_addict> I'd like to encode few clips so I can see for myself [05:01:01 CEST] <furq> ffmpeg -version | grep x265 [05:01:49 CEST] <furq> i wouldn't get your hopes up too much [05:12:46 CEST] <zumba_addict> why? [05:13:02 CEST] <zumba_addict> so x265 is still experimental? [05:13:05 CEST] <furq> it's not that much better than x264 and it's much slower [05:13:18 CEST] <zumba_addict> slower in terms of reencoding? [05:13:30 CEST] <furq> in terms of encoding and decoding [05:13:38 CEST] <zumba_addict> more cpu hungry then [05:15:49 CEST] <furq> it's about 5-10 times slower than x264 -preset veryslow if you want to see measurable gains [05:18:05 CEST] <zumba_addict> wow, ouch [05:18:17 CEST] <zumba_addict> not worth it then [05:18:39 CEST] <zumba_addict> veryslow is really slow, now with 5-10 times even slower, :D [05:21:12 CEST] <furq> it's also not widely supported yet [05:21:37 CEST] <furq> although i guess you know that since you couldn't play it [13:34:13 CEST] <Fyr> is it possible to create a video with alpha channel then merge this video with another? [13:34:54 CEST] <JEEB> yes, you can overlay things [13:52:47 CEST] <wallbroken> i tried encoding in 32 and 64 bit mode [13:52:56 CEST] <wallbroken> total different output files i get [13:53:12 CEST] <Fyr> JEEB, so it's possible to create text placed in half-transparent box and the put it onto an image? [14:05:01 CEST] <JEEB> Fyr: yes although for text I would probably use ASS subtitles with custom fonts and positioning etc [14:05:27 CEST] <Fyr> JEEB, do you have a manual for this? [14:06:19 CEST] <JEEB> look at aegisub with regards to how create the ASS and then use the ass subtitle filter [14:06:45 CEST] <JEEB> or just do it your way and have an overlay clip with alpha [14:25:00 CEST] <maziar> hi [14:26:14 CEST] <maziar> i want to stream from my hdmi to m3u8 , is it posible? **i have nginx for it [14:26:59 CEST] <maziar> how can get input from hdmi [14:27:42 CEST] <Fyr> JEEB, do you have an example for Title, like text lying towards camera? [14:31:18 CEST] <maziar> Fyr would you please kindly help me ? i want to stream from my HDMI input, is it possible ? [14:31:42 CEST] <Fyr> I've never streamed with FFMPEG. [14:37:56 CEST] <maziar> is there any body ? [14:38:26 CEST] <maziar> Fyr do you know how to record from hdmi [14:38:34 CEST] <Fyr> no [14:38:41 CEST] <Fyr> I've used only OBS. [14:38:53 CEST] <maziar> Fyr sorry what is OBS ? [14:39:14 CEST] <Fyr> https://obsproject.com/ [14:42:18 CEST] <maziar> is there any body know how to capture from HDMI [15:00:01 CEST] <iive> maziar: hdmi captures are rare and expensive. are you sure you don't want to just capture the computer screen? [15:01:05 CEST] <bencoh> blackmagic isn't *that* expensive, but ... yeah. [15:01:30 CEST] <maziar> iive yes i want to capture input from hdmi [15:03:18 CEST] <iive> maziar: do you have the hardware capture? [15:03:43 CEST] <maziar> iive what do you mean ? [15:04:18 CEST] <iive> you need some device to capture the signal. [15:04:42 CEST] <iive> you can't do it with bare hands. [15:05:35 CEST] <iive> screen capture is easier, because you take the image from the video card memory. but once the hdmi signal is out, you need a capture card. [15:09:15 CEST] <iive> if you are under linux, x11grab does the screen capture. look the manual/documentation [15:10:25 CEST] <iive> https://trac.ffmpeg.org/wiki/Capture/Desktop [15:21:45 CEST] <jpleau> hi, I can get a AVCodecParameters struct with AVStream.codecpar. How can I get the long name of the codec? [15:24:32 CEST] <durandal_1707> with that struct only? Pick codec id and then find its long name using codec_desc struct [15:25:04 CEST] <durandal_1707> that one should be global [15:25:34 CEST] <jpleau> Yes I had codec_ic, was not aware of codec_desc though [15:25:39 CEST] <jpleau> er, codec_id* [15:26:03 CEST] <jpleau> thanks durandal_1707 [16:14:38 CEST] <Fyr> guys, how does concatenation work if the framerates are different? [16:42:42 CEST] <Polochon_street> Hi ! What is the standard procedure to have both AVCodec and AVCodecContext from an opened AVFormatContext ? AVCodec first, or AVCodecContext ? [16:59:22 CEST] <cyphix> I'm using the following ffmpeg command to create gifs : ffmpeg $quiet -v warning -i lossless.mkv -i $gif_palette -lavfi "$video_options [x]; [x][1:v] paletteuse" -y $file.$ext. I would like to store this command in a file to reuse it later. But to do so, I have to remove the double quotes inside the command. But doing so, when I use it, I get "Unable to find a suitable output format for [x]. Is [x] some kind [16:59:24 CEST] <cyphix> of expanded variable? [17:06:14 CEST] <cyphix> nevermind, I found how to solve it! [17:44:17 CEST] <Polochon_street> is avcodec_open2 usable with an av_find_best_stream'd codec ? It doesn't seem to work on my case :| [18:14:03 CEST] <kdehl> Greetings. [18:17:10 CEST] <kdehl> I want to merge two video streams into one video with the original videos shown in parallel. So far so good, this is the command I use: [18:17:18 CEST] <kdehl> ffmpeg -i left.mp4 -i right.mp4 -an -y -filter_complex "pad=w=in_w*2[padded]; scale=544:408[right_scaled]; [padded][right_scaled]overlay=W/2" output.mp4 [18:17:24 CEST] <kdehl> It works well. [18:18:10 CEST] <kdehl> I just wonder whether I can replace that 544:408 with a reference to the size of the video in the first filter chain? [18:18:27 CEST] <kdehl> Pardon me if I use incorrect terminology, but I think you understand what I mean. [18:19:42 CEST] <c_14> no, filters can only access the variables of their input. not of other filters in the graph [18:20:24 CEST] <kdehl> Okay, thank you very much. [18:21:00 CEST] <kdehl> On other words, I need to make a bash script in order to automate the process. [18:25:40 CEST] <c_14> or any other scripting language, yes [18:26:47 CEST] <kdehl> Right. [18:28:06 CEST] <kdehl> So if I understand it correctly, the unnamed inputs in the filter graph are connected to the input streams specified by the -i parameter in the order that they are given? [18:29:49 CEST] <c_14> Probably, but I wouldn't depend on it. It's better to explicitly name them. [18:32:52 CEST] <kdehl> I couldn't figure out how to do that. [18:33:34 CEST] <kdehl> Never mind. [18:33:36 CEST] <kdehl> It's in the manual. [18:36:03 CEST] <kdehl> Heh, I was reading a bad guide. [19:09:45 CEST] <DelphiWorld> SUP everyone [20:40:18 CEST] <Fyr> guys, when concatenating, I see crap in the place where supposed to be video. [20:40:37 CEST] <Fyr> is it happening because of concatenation protocol? [20:41:46 CEST] <Fyr> I'm concatenating the four videos, each of them has the same encoders, sample_fmt, pix_fmt, frame rate, ar. [20:42:12 CEST] <furq> what container [20:42:15 CEST] <Fyr> however, between the second and the third video thereis only artifacts. [20:42:17 CEST] <Fyr> MKV [20:42:24 CEST] <furq> you can't use the protocol for mkv [20:42:26 CEST] <furq> use the demuxer [20:43:27 CEST] <Fyr> furq, do you mean -f concat -i my_list.txt, as mentioned in the wiki? [20:43:36 CEST] <furq> yes [20:43:46 CEST] <Fyr> it's exactly what I'm doing. [20:43:55 CEST] <Fyr> I tried MP4. [20:44:03 CEST] <Fyr> it didn't work out either. [20:46:54 CEST] <JEEB> concat in FFmpeg in various ways is a horrible hack, so depending on the container and other things it might fail. I think the concat filter might be the least bad alternative? it requires re-encoding of course [20:47:03 CEST] <JEEB> because the stuff is concatenated after decoding [20:47:24 CEST] <Fyr> I don't like re-encoding. ='( [20:47:45 CEST] <Fyr> every file in the list is played without causing troubles. [20:47:48 CEST] <Fyr> =/ [20:47:52 CEST] <JEEB> well then you always have to make sure that your parameter sets etc don't change [20:48:04 CEST] <JEEB> because otherwise you can have fun when you seek into another concatenated stream [20:48:17 CEST] <JEEB> many files only have the parameter sets either off-band or in the first IRAP [20:48:32 CEST] <JEEB> now guess what happens when you don't pass the point where the parameter sets are [20:48:43 CEST] <JEEB> and seek from file A to file B contents [20:48:43 CEST] <Fyr> artifacts [20:48:50 CEST] <JEEB> more like garbage I guess [20:49:10 CEST] <JEEB> I guess in some cases it might not be lethal [20:49:22 CEST] <JEEB> but then again in other cases and depending on the decoder they can be lethal [20:49:41 CEST] <JEEB> that is why in broadcast each IRAP contains the parameter sets [20:50:05 CEST] <JEEB> to make sure the decoder has the correct stuff at that random access point and can start playback [20:52:43 CEST] <Fyr> change of the container did the trick. [20:53:15 CEST] <JEEB> yeah, depending on container the timestamps or whatever can go woo-woo when it hits the next file [20:53:58 CEST] <JEEB> in any case, concat is something that everyone wants but that is not always easy to attain properly [20:54:22 CEST] <JEEB> thankfully I don't need the feature (yet), all my use cases have been with re-encoding so the video filter has been good enough [20:54:23 CEST] <Fyr> if somebody is interested, MPEG-TS is way better for concatenation than MKV and MP4. [20:54:42 CEST] <JEEB> yes, MPEG-TS seems to be one of the least awful alternatives when you are using the demuxer or the protocol [21:53:16 CEST] <Fyr> guys, I've just performed concatenation with concat filter and found out that it produces video of crappy quality. [21:55:03 CEST] <c_14> you probably just forgot to set video encoding options [21:55:32 CEST] <Fyr> is it possible to set stream copying? [21:55:41 CEST] <c_14> no [21:56:01 CEST] <Fyr> then how do I keep the quality? [21:56:35 CEST] <c_14> pick a decent video codec with decent quality settings [21:56:41 CEST] <c_14> like libx264 around crf 23 or so [23:09:19 CEST] <DelphiWorld> hey guys [23:09:37 CEST] <DelphiWorld> please someone try to help me with vaapi? my streams are not accepted by any streaming server... [23:10:16 CEST] <chovy> should ffmpeg process terminate when a 403 request is made for a key? [23:16:53 CEST] <DelphiWorld> can ffmpeg be a rtsp server itself, and a client send to it? [23:17:01 CEST] <DelphiWorld> sory but client receive from ffmpeg [23:23:59 CEST] <DelphiWorld> furq: here? [23:24:48 CEST] <neflux> Yo! I'm wondering how can I concat a 1 frame video generated from a still image with a video file (so that the image would be the first frame of the output video)? [23:26:57 CEST] <DelphiWorld> this one work perfectly: [23:26:57 CEST] <DelphiWorld> ffmpeg -i udp://@239.100.0.2:1234 -acodec libfdk_aac -ab 64k -vcodec libx264 -vb 1000k -s 640x360 -f rtsp -rtsp_transport tcp rtsp://197.140.2.6/live/almajd03 [23:28:30 CEST] <Polochon_street> what is the difference between AVCodecParameters and AVCodecContext ? [23:32:10 CEST] <neflux> delphiWorld Ok I'm lazy but not stupid that looks too shady to me hahah [23:32:35 CEST] <DelphiWorld> neffwhat you're talking about? :-P [23:33:41 CEST] <neflux> I guess there were desperate enough people here that actually tried everything you gave to them hahah [23:34:05 CEST] <DelphiWorld> LOL [23:34:18 CEST] <DelphiWorld> my vaapi is working now [23:34:19 CEST] <DelphiWorld> with rtsp [23:34:21 CEST] <DelphiWorld> fuck rtmp :P [23:34:57 CEST] <neflux> ffmpeg -loop 1 -framerate 24 -t 1 -i "image.jpg" -t 1 -f lavfi -i aevalsrc=0 -i "input.webm" -filter_complex '[0:0] [1:0] [2:0] [2:1] concat=n=2:v=1:a=1' "output.webm" ///// is giving me No such filter: '[0:0]' [23:38:56 CEST] <neflux> Maaan I'm just trying to change the thumbnail and I think this is the only way. Because 4chan uses the first frame to generate it. [00:00:00 CEST] --- Mon Aug 1 2016

1 0