[Ffmpeg-devel-irc] ffmpeg.log.20140605

burek burek021 at gmail.com
Fri Jun 6 02:05:01 CEST 2014


[00:04] <llogan> synth_: yeah, nobody has tried it yet and i added FFmpeg to it somewhat recently
[00:05] <synth_> ohhh
[00:05] <synth_> brb gotta reboot
[00:38] <synth_> back
[01:42] <a1fa> hello, i am trying to convert bunch of dv files to mpeg to burn to pal-dvd, is there anything special i should use when ffmpeg convering them?
[01:42] <c_14> If you're focusing on pal-dvd you can use -target pal-dvd
[01:43] <a1fa> and it has default bitrate and etc?
[01:43] <c_14> bitrate, codecs, buffer sizes, the whole shebang
[01:43] <a1fa> shweeet
[01:43] <a1fa> ffmpeg -y -i input.dv -target pal-dvd output?
[01:44] <c_14> should be fine
[01:44] <a1fa> coolio
[01:44] <synth_> bah
[01:45] <a1fa> i tried to dvigrab them directly into pipe ffmpeg
[01:45] <a1fa> but ffmpeg was crashing like a drunk teenager
[01:46] <a1fa> grabbing from firewire camcorder
[01:48] <fflogger0> a1fa, please use a pastebin site (like www.pastebin.com) to show your ffmpeg command and the complete console output.
[01:49] <klaxa> fflogger0 seems outdated and redundant
[01:50] <c_14> Been here since yesterday, apparently. Not sure why.
[01:52] <a1fa> old athlons suck for ffmpeg ;(
[01:52] <klaxa> s/athlons/computers/
[01:52] <a1fa> 387M dv file to pal-dvd took user    1m5.485s
[01:53] <a1fa> input     Stream #0:0: Video: dvvideo, yuv411p, 720x480 [SAR 32:27 DAR 16:9], 28771 kb/s, 29.97 tbr, 29.97 tbn, 29.97 tbc
[01:53] <a1fa> output     Stream #0:0: Video: mpeg2video, yuv420p, 720x576 [SAR 64:45 DAR 16:9], q=2-31, 6000 kb/s, 90k tbn, 25 tbc
[01:53] <a1fa> is that going to look wierd?
[01:54] <klaxa> try to play it back, if it doesn't look weird, it doesn't look weird
[01:55] <a1fa> no looks fine
[01:55] <a1fa> on 720p.. but they are taking this to europe.. not sure if they have 720p over there
[01:56] <a1fa> so i bet it will get stretched to fit their screen
[01:56] <a1fa> anyone know of a vhs player with firewire?
[01:56] <sacarasc> 720x576 is not 720p, it's 576p, if progressive.
[01:56] <a1fa> i need to convert old family tapes to dvd
[01:56] <a1fa> sacarasc: i played it on a computer with 720p output
[01:57] <c_14> the output you're creating is anamorphic with a dar of 16:9
[01:57] <a1fa> good or bad?
[01:57] <sacarasc> And over here, we only have 4k, we don't do lower.
[01:58] <a1fa> where is there?
[01:58] <klaxa> europe
[01:58] <a1fa> lol
[01:58] <sacarasc> Yes.
[01:58] <a1fa> i wonder if i should burn these mpegs as dvd format, or burn them as DVD-File
[01:59] <a1fa> %Cpu0  : 65.2 us,  5.6 sy,  4.6 ni, 23.8 id,  0.3 wa,  0.3 hi,  0.0 si,  0.0 st
[01:59] <a1fa> %Cpu1  : 61.3 us,  5.3 sy, 10.3 ni, 22.5 id,  0.3 wa,  0.3 hi,  0.0 si,  0.0 st
[01:59] <a1fa> blah
[02:03] <a1fa> well, i just effed up on the conversion, and need to start the capture all over
[02:03] <a1fa> i am going to try dvgrab into ffmpeg
[02:05] <a1fa> looks like its not possible
[02:05] <a1fa> to do a dvgrab
[02:15] <thebombzen> a1fa: FFmpeg supports dv1394 FireWire grabbing
[02:15] <thebombzen> directly from FFmpeg
[02:54] <llogan> a1fa: dvgrab to ffmpeg should work fine: dvgrab -size 0 -noavc - | ffmpeg -i - <other junk> output
[02:54] <llogan> or something like that
[02:55] <llogan> or try the suggestion from thebombzen. or just grab the files with dvgrab and encode those. that will give you the advantage of being able to re-encode if you screw up.
[06:11] <bornpilot> so doing some work with mpegts segments what is a global header?
[06:13] <bornpilot> I see it the docs with ssegments or stream_segments but what are they?
[07:00] <bornpilot> so doing some work with mpegts segments what is a global header?
[07:00] <bornpilot> 11:13
[07:00] <bornpilot> I see it the docs with ssegments or stream_segments but what are they?
[07:01] <bornpilot> have you contribted to an open source project?
[07:46] <Guest2333> Hi, I'm trying to understand how to get truepeak info from ffmpeg. cmd: ffmpeg -nostats -i A00010-01.wav -filter_complex "ebur128" -f null -     this works, but I'm not sure how to add truepeak to the output?
[11:07] <fresh1> Hi, I'm using ffmpeg for real-time encoding and want to get thumbnail from input-video stream every period of time. I've tried to use thumbnail filter with parameter n =ª  50 (thumbnail=50). And on my cpu (Intel(R) Xeon(R) CPU @ 3.47GHz stepping 02) using this filter increases average cpu usage for 5%-10% --- this very high value I thinª k. Can you suggest me how can I reduce CPU usage and achieve the
[11:07] <fresh1> same goal? My global filter looks like: [#0x110]yadif,split=2[vidin][thumbin];[vidin]scale=1920:1080[vª idout];[thumbin]thumbnail=50[thumbout] gop size 75 and I'm using preset "fast", high profile; MP2 --> H.264
[11:09] <fresh1> Sorry, copy-pasted from vim...  Hi, I'm using ffmpeg for real-time encoding and want to get thumbnail from input-video stream every period of time. I've tried to use thumbnail filter with parameter n = 50 (thumbnail=50). And on my cpu (Intel(R) Xeon(R) CPU @ 3.47GHz stepping 02) using this filter increases average cpu usage for 5%-10% --- this very high value I think. Can you suggest me how can I reduce CPU
[11:09] <fresh1> usage and achieve the same goal? My global filter looks like: [#0x110]yadif,split=2[vidin][thumbin];[vidin]scale=1920:1080[vidout];[thumbin]thumbnail=50[thumbout] gop size 75 and I'm using preset "fast", high profile; MP2 --> H.264
[11:20] <fresh1> quit
[11:55] <thshdw> can ffmpeg create a master_playlist.m3u8 from variant_playlists.m3u8? I'm attempting to do this through this command http://pastebin.com/3if48fPm but its only segmenting variant_500k. the other variant variant_150k does not get segmented.
[15:13] <termos> thshdw: don't think so, i'm currently making a master playlist myself and I ended up creating it myself
[15:14] <termos> it does make the Manifest file for smoothstreaming though, which is nice
[17:29] <obelix666> hi what is the shortest clip that ffmpeg can transcode? I am getting inconsistent results for clips around the 50ms mark and just wondering if there is a hard lower limit.
[17:30] <JEEB> there shouldn't be
[17:30] <JEEB> I mean, one sample is the minimum I'd guess
[17:30] <JEEB> and then there's the exceptions where someone was lazy with a third party library wrapper
[17:30] <JEEB> I think the libx265 wrapper doesn't output anything if you don't give it enough pictures
[17:31] <obelix666> sometimes i get only an audio track out and sometimes only a video track out
[17:31] <obelix666> using libx264
[17:33] <JEEB> the libx264 wrapper should handle even a single picture just fine
[17:34] <JEEB> unless someone broke it lately
[17:35] <JEEB> in other words, as long as something gets decoded/read, it should also get output (except for possible exceptions, but those should be rather rare)
[17:41] <kmacleod> obelix666, JEEB, I've just been doing something like that and x264 is telling me it has read 36 input frames before handing me any
[17:41] <kmacleod>     Stream #0:0, 36, 1/90000: Video: h264 (High), yuv420p, 720x480, 1/60, 30 fps, 30 tbr, 90k tbn, 60 tbc
[17:41] <JEEB> yes, but it should still give them all to you
[17:42] <JEEB> and it depends on the settings
[17:42] <JEEB> as well as if there is no more frames it should still work
[17:42] <JEEB> only the libx26_5_ wrapper should be broken with this
[17:42] <JEEB> not the libx264 one
[17:44] <kmacleod> avformat_find_stream_info() seems to be the one eating all the initial frames.  there may be a way to pull the processed frames but I haven't looked for it.
[17:44] <JEEB> no need for that
[17:44] <Fjorgynn> this ffmpeg tool is awesome
[17:45] <Fjorgynn> you should try it
[17:45] <JEEB> it will then start from the beginning of the buffer after it has probed the stream
[18:08] <thshdw> @termos Did you make each stream individually? Or were you able to make you multiple streams off one command line command?
[18:38] <ChocolateArmpits> Is it possible to seek if frames, rather than timecode, in ffmpeg without segmenter ?
[19:14] <BtbN> Is it possible to take 5 videos, and "interleave" their frames? So that when i have the 5 videos "a0, a1, a2, ...", "b0, b1, b2, ...", ...  i end up with a single video with a frame order of "a0, b0, c0, d0, e0, a1, b1, c1, d1, e1, a2, ..."?
[19:29] <llogan> ChocolateArmpits: you can try the select filter. maybe something like select='gte(n\,1234)'
[19:35] <BtbN> Hm, seems like that's not possible at all
[19:36] <BtbN> maybe with a crazy avisynth script, but that seems to be windows-only
[20:05] <ChocolateArmpits> BtbN, it should be possible if you first output all your videos  as an image sequence to separate folders and then perform a renaming procedure so that the frames will fall into their place all due to their names.
[20:12] <theekoz> on my mac Say I'm trying to have my input video on my second drive, it won't let me run the commandline when I drag the file into terminal
[20:12] <pzich> I missed that BtbN was someone's nick, and was trying to figure out what the hell that stood for
[20:12] <theekoz> anyone know the right command to access from a second HD
[20:13] <pzich> /Volumes/<drivename/... ?
[20:13] <theekoz> yea
[20:13] <theekoz> that's what's not working
[20:13] <llogan> you can drag files into the os x terminal?
[20:13] <theekoz> yes
[20:13] <theekoz> turns it into the path
[20:13] <llogan> blasphemy.
[20:17] <llogan> BtbN: "-filter_complex framepack=frameseq" might be worth a try
[20:23] <llogan> BtbN: or maybe http://ffmpeg.org/ffmpeg-filters.html#interleave_002c-ainterleave
[20:41] <dogitu> is there a reason that ffmpeg would print a partial x264 build number (e.g. x264 core 142 rather than x264 0.142.2431) in it's .mkv output (on linux)?
[20:49] <ChocolateArmpits> llogan: What does the "n\" stand for? Currently evaluated frame?
[20:50] <llogan> http://ffmpeg.org/ffmpeg-filters.html#select_002c-aselect
[20:50] <llogan> The (sequential) number of the filtered frame, starting from 0.
[20:51] <ChocolateArmpits> Oh strange, overlooked that
[20:51] <ChocolateArmpits> thanks
[20:55] <rsevero> Hi. How can I set a alsa virtual device as input for ffmpeg? By "virtual device" I mean the ones defined in .asoundrc for example. The ones listed by "arecord -L" not the ones listed by "arecord -l".
[20:58] <pzich> http://www.ffmpeg.org/ffmpeg-devices.html#alsa ?
[20:58] <rsevero> Well, it's simple after all. Just use "plug:DEVICE_NAME".
[20:59] <rsevero> pzich: I can't find the info I want on the page you sent but I believe the solution I just posted works. Thanks anyway.
[20:59] <llogan> https://trac.ffmpeg.org/wiki/Capture/ALSA has some additional into
[20:59] <llogan> *info
[21:04] <PoWeRKiLL> Hi
[21:04] <PoWeRKiLL> How can I detect the presence of an image in a video ?
[21:05] <pzich> an image?
[21:07] <PoWeRKiLL> like a logo in a video
[21:13] <rsevero> PoWeRKiLL: take a look at the OpenCV library.
[21:17] <PoWeRKiLL> Yes I'm looking on it I was asking myself if someone can point me out
[21:20] <BtbN> ChocolateArmpits, that would propably be too big for any of my disks to hold
[21:20] <BtbN> each h264 video alone is several GB in size
[21:21] <BtbN> llogan, hm, interleave would propably shuffle the frames, as they have the same timestamp each time
[21:25] <BtbN> also: [AVFilterGraph @ 0000004A734D5DA0] Too many inputs specified for the "interleave" filter.
[21:27] <llogan> that means little without any context
[21:27] <BtbN> well, 5 inputs is simply too much for the interleave filter
[21:28] <BtbN> framepack with frameseq also is quite hardcoded for two input videos
[21:28] <llogan> did you set nb_inputs for interleave?
[21:29] <BtbN> oh, missed that option
[21:30] <llogan> this is an example of why we are always so insistant on seeing the command and complete console output instead of just a segment of the console output
[21:30] <pzich> in general, it's best to paste (not here, use pastie.org or something) the whole command you ran and the output from ffmpeg.
[21:30] <BtbN> Well, now it seems to be doing something
[21:31] <BtbN> no error yet, it's eating up ram and cpu
[21:31] <BtbN> no output at all so far
[21:31] <ChocolateArmpits> llogan: select filter doesn't seem to work with frame numbers in hundred thousands. Many of the videos I deal with are around 2 hours long, sometimes longer. Specifying frame number of 233,000 gave me an error "Too large number of skipped frames 223000 > 60000"
[21:31] <BtbN> i hope it doesn't try to decode all 5 videos in ram
[21:31] <llogan> BtbN: just encode a few seconds with -t. it should be ovious if it works as expected or not
[21:32] <llogan> ChocolateArmpits: i don't know what else to recommend for that
[21:34] <ChocolateArmpits> I found suggestions to use pkt_dts_time from ffprobe for frame times, but on some videos it's adjusted by "start time" which makes the number not the duration number
[21:34] <BtbN> this is strange. i cut the input clips a little shorter, but the output file keeps growing and growing, way bigger than the input file sizes combined.
[21:34] <llogan> ChocolateArmpits: try the mailing list i guess
[21:34] <BtbN> That's the command: ffmpeg -i cam5.avi -i cam4.avi -i cam3.avi -i cam2.avi -i cam1.avi -filter_complex "[0:v][1:v][2:v][3:v][4:v] interleave=n=5" -t 1 out.mp4
[21:34] <ChocolateArmpits> BtbN: that's because there are more frames
[21:34] <ChocolateArmpits> combined
[21:35] <pzich> 5x, it seems
[21:35] <BtbN> hm? combinding 4 1MB files should never generate a 500MB output file
[21:35] <BtbN> -d
[21:35] <pzich> is it actually that extreme?
[21:35] <BtbN> *5 1MB files
[21:35] <BtbN> yes
[21:35] <ChocolateArmpits> You didn't specify encode settings it seems
[21:36] <ChocolateArmpits> What does it say in the output?
[21:36] <BtbN> no output at all
[21:36] <ChocolateArmpits> There should be something, there "Input" section and then there's "Output" section in the CLI
[21:36] <pzich> not even the copyright and config stuff?
[21:36] <BtbN> nope, nothing
[21:36] <pzich> and that's the complete command you're running?
[21:36] <BtbN> yep
[21:37] <pzich> did you by chance alias ffmpeg to add silent flags?
[21:37] <BtbN> no, running ffmpeg alone or cutting the files shorter worked fine
[21:38] <BtbN> If i set the loglevel to verbose i get this: http://bpaste.net/show/mxHvPnRKDOH3zxtnoM9f/
[21:38] <BtbN> stopping exactly where the paste stops
[21:40] <ChocolateArmpits> ""Output #0, mp4, to 'out.mp4':   Metadata:     encoder         : Lavf55.33.100     Stream #0:0: Video: h264 (libx264) ([33][0][0][0] / 0x0021), yuv420p, 1920x1080 [SAR 1:1 DAR 16:9], q=-1--1, 1000k tbn, 1000k tbc (default)"
[21:40] <ChocolateArmpits> this is your output
[21:40] <BtbN> yeah, looks ok except for it never stopping to fill it
[21:43] <BtbN> i think i'll have to write a small programm to do this
[21:44] <llogan> why do you want to show the frames like this?
[21:44] <BtbN> It's for an autostereoscopic 3D display which combines 5 camera angles
[21:45] <ChocolateArmpits> if this isn't working out for you, as per my previous suggestion, output each video to an image sequence and write a bash script with two for loops to rename the image sequences
[21:45] <BtbN> Well, the full video is 2 hours long
[21:45] <ChocolateArmpits> then combine the whol sequence into a video with ffmpeg again
[21:46] <ChocolateArmpits> of each camera?
[21:46] <BtbN> yep
[21:46] <llogan> someone has probably done this with avisynth
[21:47] <ChocolateArmpits> BtbN previously said he's no on windows
[21:47] <ChocolateArmpits> not*
[21:47] <BtbN> I'm at home now, i have windows here, but the combine script should ultimately run on a linux box
[21:48] <llogan> works fine with wine
[21:49] <BtbN> hm, i don't think that would get accepted
[21:49] <ChocolateArmpits> Do you need the final video same length (2 hours)?
[21:50] <BtbN> yeah, the display is running at 5 times 24/30/60 hz
[21:52] <llogan> there is avxsynth for linux, but i know nothing of it, or if it is stil active, or even usable
[21:52] <JEEB> it was posted and left there to wither
[21:52] <JEEB> basically it ran that one thing that was never released in public by netflix or whatever that did the porting of the old version of avs
[21:52] <JEEB> and that's it
[21:52] <JEEB> vapoursynth is much better for video if you need something like avs that is cross-platform
[21:53] <JEEB> or just run avs 2.6.x under wine
[23:09] <audio_question> hi guys! i managed to "encrypt" a mpeg1 video file by xoring some pixels with a value (etc...) and i want to encrypt as well the audio part of that video. can u tell me what is the source from libavcodec that i should modify in order to edit audio mp2 encoding/decoding?
[23:15] <audio_question> hi guys! i managed to "encrypt" a mpeg1 video file by xoring some pixels with a value (etc...) and i want to encrypt as well the audio part of that video. can u tell me what is the source from libavcodec that i should modify in order to edit audio mp2 encoding/decoding? can any1 help me pls?
[23:50] <llogan> JEEB: i see. i can never remember the details of all of the variants
[00:00] --- Fri Jun  6 2014


More information about the Ffmpeg-devel-irc mailing list