[Ffmpeg-devel-irc] ffmpeg.log.20161004

Wed Oct 5 03:05:01 EEST 2016

[00:06:42 CEST] <Wilawar> In summary, normal video but long audio cut-outs on VLC, shorter audio cut-outs and video slowdowns with MPC-HC and almost unnoticable audio problems but still video slowdowns with FFplay. Sorry but I think its a problem with FFmpeg
[00:07:20 CEST] <c_14> open an issue on trac?
[00:08:39 CEST] <Wilawar> If Im the only one to be able to reproduce it, Im not sure thats the way to go yet. Id feel better with confirmation from someone that its not something Ive been dumb about
[00:21:23 CEST] <NobCat> hi guys
[00:21:45 CEST] <NobCat> anyone around to help me recover some video?
[00:21:48 CEST] <NobCat> http://pastebin.com/sJRLG3qe
[00:26:07 CEST] <Polochon_street> is there any way to obtain the specific error of a failed avcodec_open2 error? I passed a perfectly fine codec to it ("PCM signed 16-bit little-endian"), but it fails :\
[00:26:18 CEST] <Polochon_street> (the same code works with every of my other audio files)
[00:29:27 CEST] <Polochon_street> besides, the previous avcodec_alloc_context() seemed to have worked fine :\
[00:31:33 CEST] <Wilawar> Another thing I tried: Opening it in Audacity 2.1.0. I do remember installing some FFmpeg codec thing for it, I dont know whether it has been used, but the audio is fine as far as I can tell
[01:33:20 CEST] <qmr> https://www.youtube.com/watch?v=3GJUM6pCpew  may be relevent to your interests
[02:34:22 CEST] <SchrodingersScat> qmr: that is interesting actually
[02:42:39 CEST] <klaxa> >ntcs: not the smartes choice
[02:42:39 CEST] <klaxa> heh
[02:42:43 CEST] <klaxa> true
[06:51:36 CEST] <wallbroken> hi
[08:42:40 CEST] <wallbroken> is there any important difference between mp4 and mov?
[08:45:47 CEST] <sunny26> Hi I have added a support for a hardware decoder in ffmpeg version 2.8 as I have to hook it up with chroimum, how can I upstream this changes to version 2.8  of ffmpeg?
[08:46:17 CEST] <sunny26> latest version is 3.1.2, do support added in 2.8 will be upstreamed to 3.1.2 !
[08:46:40 CEST] <sunny26> I am new to upstreaming and its SOP, so please someone help
[09:35:39 CEST] <c_14> sunny26: clone ffmpeg git, rebase your patch on master head, send the rebased patch to the ffmpeg-devel mailing list https://ffmpeg.org/developer.html
[09:37:33 CEST] <sunny26> <c_14>: ok thanks
[09:39:22 CEST] <Spring> I read on Bugzilla that Chromium uses ffmpeg internally for decoding, anyone know if this is correct?
[09:39:47 CEST] <Spring> surprisingly google shows exactly 0 results for 'ffmpeg chromium'
[09:40:48 CEST] <spaam> i found several
[09:41:03 CEST] <sunny26> <Spring> Suprisingly it exit here, https://chromium.googlesource.com/chromium/third_party/ffmpeg.git
[09:41:46 CEST] <Spring> sunny26, where did you find the link btw?
[09:42:01 CEST] <spaam> google: "ffmpeg chromium"
[09:42:10 CEST] <spaam> second hit
[09:42:11 CEST] <sunny26> google.
[09:42:23 CEST] <Spring> I have, that result doesn't display
[09:43:00 CEST] <Spring> https://a.uguu.se/llP5LCQkql5O.png
[09:43:16 CEST] <sunny26> I think we discuss ffmpeg related problems and issuses here, not related to google search results.
[09:43:51 CEST] <sunny26> anyways the link I provided is accurate
[09:44:02 CEST] <Spring> I realize that, it was showing that I can't find this myself using the same query is all
[09:44:20 CEST] <Spring> or anything related for that matter
[09:44:33 CEST] <sunny26> Ya that's fine
[09:46:31 CEST] <sunny26> searching 'ffmpeg chromium' gives the link i sent you in second result, check this https://www.google.co.in/search?client=ubuntu&channel=fs&q=ffmpeg+chromium&ie=utf-8&oe=utf-8&gfe_rd=cr&ei=Cl7zV-CqCO_x8AeBxpOwDQ
[09:48:27 CEST] <Spring> odd. After relaunching Firefox the same query displays the results.
[11:17:43 CEST] <oakwhiz> Question: Can this type of video be made using filters in ffmpeg? https://www.youtube.com/watch?v=P5g-lCsOoNs
[11:19:10 CEST] <oakwhiz> I was thinking of using setpts, a colorkey, and an overlay, and that would get 1 copy of the same video but delayed. The recursion is confusing however
[11:20:20 CEST] <oakwhiz> Extending it to an arbitrary number of copies would seem to require an ever-expanding command line
[11:59:01 CEST] <Bernmeister> I've read https://trac.ffmpeg.org/wiki/Seeking and am confused by the first (of three) example in "Cutting small sections".  Why does "-ss 00:01:00 ... -to 00:02:00" yield a total duration of 3 minutes rather than 2 minutes (as per the other two examples)?  I have tested this myself and indeed I get three minutes...but why?
[12:01:56 CEST] <furq> it doesn't
[12:05:17 CEST] <n4zarh> I need to be able to add adpcm_ima_oki encoder to my project, but ffmpeg allows only for decoding this format. Is there any way to add that decoder to list? If not, can anyone tell me what codec should I look for?
[12:06:00 CEST] <c_14> furq: Bernmeister, it does when -ss is an input option because using -ss as an input option adjusts the timestamps by that amount (unless -copyts is used)
[12:06:24 CEST] <furq> that wouldn't give a duration of three minutes though
[12:06:31 CEST] <furq> it'd cut from 1:00 to 3:00
[12:06:41 CEST] <c_14> eh, yeah
[12:06:43 CEST] <c_14> right
[12:06:53 CEST] <furq> n4zarh: it doesn't look like there's an encoder for that format
[12:07:06 CEST] <c_14> Arithmetic is hard
[12:07:26 CEST] <n4zarh> yeah, that's why I ask, I'm not 100% sure but I might need that encoder
[12:07:35 CEST] <furq> Bernmeister: the other two examples give a duration of 1:00, the first one is 2:00
[12:07:59 CEST] <furq> if that's what you meant then it's explained just above the examples
[12:19:30 CEST] <Bernmeister> c_14, furq: Thanks - I've run a few tests just now and I suppose the simplest solution for me is to use the -t option in lieu of -to when -ss is used as an input option.  Otherwise I get audio but no video at the start OR I get more time/duration that I expect.
[12:35:47 CEST] <n4zarh> I seem to have problem with filling audio frame while trying to encode audio, does anyone know something about that function?
[12:38:44 CEST] <n4zarh> I try to get data from mic on android, encode it and send it to ip camera; definitely tries to encode data for adpcm codecs (but they are not what I need), but fail with -22 error code on pcm_mulaw/alaw encoders, not sure what am I missing
[12:39:54 CEST] <BtbN> send audio to an IP camera? oO
[12:40:28 CEST] <n4zarh> it has some kind of speaker
[12:41:29 CEST] <n4zarh> it can record video and audio, but also it can play audio sent with it's API
[12:43:33 CEST] <n4zarh> so I need to encode PCM data from my smartphone microphone to pcm_mulaw codec (since camera uses this one); and that's where I fail, I can't encode any frame, because my code crashes on avcodec_fill_audio_frame function
[12:46:54 CEST] <n4zarh> or actually av_samples_get_buffer_size fails
[12:56:08 CEST] <n4zarh> I think I've got reason for error - why frame_size for pcm_mulaw context equals 0?
[12:56:44 CEST] <n4zarh> it has specific sizes for every ADPCM codec I tried using, but it's always 0 for alaw/mulaw
[13:15:01 CEST] <Mavrik> hmm :/
[13:15:13 CEST] <Mavrik> It's usually worth looking at the source in these cases
[15:03:28 CEST] <wallbroken> what's the diff between mp4 and mov?
[15:05:06 CEST] <BtbN> the file extension
[15:07:21 CEST] <ritsuka> and some things are a bit different in the file format
[15:07:30 CEST] <iive> mp4 is standard based on apple mov
[15:16:05 CEST] <JEEB> but there are things defined differently between 14496-12 and qtff.pdf
[15:16:07 CEST] <wallbroken> not clear answers
[15:16:13 CEST] <JEEB> so while they are similar, they are not the same
[15:16:29 CEST] <wallbroken> mp4 does not accept some metadata that mov does
[15:16:35 CEST] <JEEB> mov is what was used as the base for ISOBMFF (which is called "mp4" colloquially)
[15:16:44 CEST] <JEEB> not surprising :P
[15:16:51 CEST] <JEEB> mov was never standardized
[15:17:06 CEST] <JEEB> ISOBMFF is actually standardized so you shouldn't be able to put random crap in it
[15:17:19 CEST] <JEEB> also *how* those tags are put there can differ
[15:17:43 CEST] <wallbroken> JEEB, mp4 is standard or apple proprietary?
[15:17:50 CEST] <JEEB> ISOBMFF is an ISO standard
[15:18:01 CEST] <JEEB> they took apple's QT File Format and made it intot a standard file format
[15:18:25 CEST] <JEEB> QT File Format also has an Apple document of course, which is why I said qtff.pdf :P
[15:18:33 CEST] <JEEB> you can think of MOV as Apple's AVI
[15:18:42 CEST] <wallbroken> ffmpeg -i input.mov -vf "hflip,vflip,scale=1024:-2" -map_metadata 0 -metadata:s:v rotate="" -c:a copy output.mov
[15:18:43 CEST] <wallbroken> it's ok
[15:18:47 CEST] <wallbroken> ffmpeg -i input.mov -vf "hflip,vflip,scale=1024:-2" -map_metadata 0 -metadata:s:v rotate="" -c:a copy output.mp4
[15:18:51 CEST] <wallbroken> gives to me an error
[15:21:39 CEST] <furq> were you planning on sharing the error with us
[16:46:51 CEST] <transhuman> hi is there a sample program anywhere anyone knows about that works from a local web server that uses ffmpeg to translate from different formats like raw to mp3 and mp3 to raw?9(for example)
[16:47:44 CEST] <transhuman> also recording the formats would be a nice part of it if I could have such a thing...I am trying to work on apache with audio
[16:50:21 CEST] <ozette> is there any documentation about -skip_frame nokey ?
[16:51:08 CEST] <furq> https://www.ffmpeg.org/ffmpeg-codecs.html#Codec-Options
[16:51:12 CEST] <furq> ozette: it's listed in there
[17:06:53 CEST] <ozette> furq: thansk
[19:18:42 CEST] <Check_> Hello
[19:21:54 CEST] <pgorley> hi, is there a way to disable vaapi x11 but not vaapi drm?
[19:34:01 CEST] <pgorley> nvm, found my answer, if you configure with disable-xlib and enable-vaapi, drm is enabled and x11 isn't
[20:50:56 CEST] <cowai> Anybody know why the bitrate is ever increasing throughout the render when using CRF?
[20:51:29 CEST] <BtbN> Because the video is getting more complex.
[20:51:31 CEST] <cowai> 24minutes in, its triple the bitrate as it was the first five minutes.
[20:51:40 CEST] <cowai> Its never going down
[20:51:52 CEST] <furq> that's the average for the whole file
[20:51:55 CEST] <cowai> Ah
[20:52:05 CEST] <furq> if the first two minutes are credits or something then it'll keep increasing
[20:52:07 CEST] <cowai> So its the average.
[20:52:10 CEST] <cowai> yes.
[20:52:13 CEST] <cowai> That explains it
[20:53:02 CEST] <furq> a rolling average would be nice
[20:53:23 CEST] <furq> right now it's just size / duration
[20:53:40 CEST] <cowai> Average last 10 sec or something would be nice.
[21:31:13 CEST] <teratorn> how would one go about generating a testsrc video of, say 9 seconds in length 30fps, but split in to precisely 3 different video files of 90 frames each?
[21:31:41 CEST] <teratorn> I guess I can just generate it, then cut it? but I'm not sure how, and not sure about keyframe boundaries being a problem...
[21:32:40 CEST] <llogan> you could use trim filter
[21:33:43 CEST] <teratorn> llogan: I'll check it, thank you!
[21:36:03 CEST] <DHE> segment output?
[21:36:28 CEST] <llogan> that too probably
[21:36:28 CEST] <DHE> so many ways, it makes me feel sorry for the cat
[21:39:51 CEST] <llogan> teratorn: you'll probably need to add setpts=PTS-STARTPTS after each trim
[21:39:59 CEST] <llogan> IIRC
[21:44:21 CEST] <teratorn> I can just use frame numbers, not timestamps
[21:45:57 CEST] <AlVal> freebsd, getting ERROR: opencv-core not found using pkg-config
[21:47:09 CEST] <AlVal> using portmaster to upgrade from 2.8.8_2,1 to 2.8.8_3,1
[21:48:15 CEST] <AlVal> interestingly opencv2-core just had an update too
[21:51:13 CEST] <teratorn> llogan: so I figured out how to do this using 3 different invocations of ffmpeg with different trim parameters... but I suppose I can use a complex filtergraph to do it in one invocation, right? :)
[21:51:49 CEST] <teratorn> ie, one testsrc input, 3 -vf trim outputs?
[21:52:42 CEST] <llogan> teratorn: ffmpeg -y -f lavfi -i testsrc=d=9:r=30 -filter_complex "[0:v]trim=start_frame=0:end_frame=90,setpts=PTS-STARTPTS[v0];[0:v]trim=start_frame=90:end_frame=180[v1];[0:v]trim=start_frame=180:end_frame=270,setpts=PTS-STARTPTS[v2]" -map "[v0]" out0 -map "[v1]" out1 -map "[v2]" out2
[21:53:00 CEST] <llogan> assuming you actually want 30 frame rate instead of ntsc
[21:53:48 CEST] <llogan> you need to setpts because trim doesn't reset the timestamps
[21:54:28 CEST] <llogan> unless you want it to do that...
[21:54:48 CEST] <teratorn> oh, so my files that I created with separate invocations of trip have non-zero starting pts's.. the 2nd and 3rd files that is... I guess most players don't care.. :/
[21:54:54 CEST] <teratorn> s/trip/trim/
[21:55:26 CEST] <teratorn> llogan: awesome, thanks
[21:55:32 CEST] <llogan> you can check with "ffprobe -v error -show_frames input"
[21:55:52 CEST] <teratorn> curious.. I built ffmpeg from source but I did not get an ffplay binary...
[21:56:00 CEST] <llogan> it requires SDL2
[21:56:07 CEST] <teratorn> ah, must have forgot that options
[21:56:55 CEST] <llogan> http://ffmpeg.org/index.html#sdl1
[21:57:16 CEST] <llogan> ^ SDL1 support was dropped recently in favor of SDL2
[21:57:35 CEST] <teratorn> SDL2 doesn't even support the nice YUVOverlay thingo that SDL1 has
[21:57:46 CEST] <teratorn> or at least it didn't, last I checked... :/
[21:57:53 CEST] Action: llogan has no clue about SDL*
[21:58:11 CEST] <teratorn> would be educational to see how ffplay utilizes SDL2 for rendering yuv and/or rgb frames
[22:01:45 CEST] <llogan> AlVal: did you install opencv-core?
[22:02:39 CEST] <llogan> alternatively, use OPENCV=off. but i'm going to blame your build script/ports system unless you can show that it is a ffmpeg problem
[22:03:11 CEST] <AlVal> llogan:  The graphics/opencv-core port moved to graphics/opencv2-core         ===>>> Reason: Renamed
[22:03:44 CEST] <llogan> seems like a freebsd issue
[22:04:13 CEST] <AlVal> ok so its not like thats something that ffmpeg hardcodes or anything
[22:04:47 CEST] <AlVal> but rather the os or ports management system should be responsible for maintaining the references to
[22:06:20 CEST] <llogan> you have to explicitly tell ffmpeg you want to --enable-libopencv. how you provide the required headers is up to you/your package management system
[22:07:35 CEST] <AlVal> llogan: thanks
[22:07:51 CEST] <teratorn> llogan: perfect, you just forgot the setpts for the middle video ;)
[22:08:38 CEST] <llogan> inconveivable. must have been packet loss.
[22:09:04 CEST] <teratorn> and indeed, ffplay did not start playing frames until the pts said so... it doesn't care what the first frame's pts is I guess... I think other players are perhaps more forgiving
[22:09:27 CEST] <teratorn> even though it rendered the first frame immediately...
[22:09:51 CEST] <teratorn> which is cool, I guess. it's made for testing purposes, mostly, I guess.
[22:20:49 CEST] <teratorn> hmm, is there any problem with having a huge gap in timestamps for a h264 video stream that is nominally 30fps, for example?
[22:21:14 CEST] <teratorn> i.e. will players just sit on the last displayed frame during the gap?
[22:21:46 CEST] <teratorn> use-case is overlaying a static image on top of a given frame and "pausing" there for a few seconds
[22:42:58 CEST] <kepstin> teratorn: assuming you're storing the video in a container that supports vfr, that should work fine. And you can have ffmpeg duplicate the frame to fill in the gap by using the fps filter if you prefer.
[22:43:26 CEST] <teratorn> kepstin: mp4
[22:45:34 CEST] <teratorn> kepstin: but what about "pausing" the input video at a given frame... duplicating that frame with fps filter *and* overlaying another stream during that pause period (the static frame duplicated with the fps filter) ? ;)
[22:46:20 CEST] <teratorn> I don't mind writing code, if the ffmpeg command line can't quite handle this... :/ right now I'm trying to figure out if it's possible from the command-line or not
[22:46:47 CEST] <teratorn> I guess one approach would be to "steal" a frame from the input video, at the right offset where I want the pause/overlay to happen
[22:47:02 CEST] <teratorn> ie in a separate ahead-of-time step
[22:47:05 CEST] <kepstin> teratorn: if you put the fps filter after everything else, it'll convert the video to constant framerate by duplicating frames. You can do the "pause" with careful use of the setpts filter, I guess. be tricky to time the overlay but it should be doable.
[22:47:56 CEST] <kepstin> like many things, it can be done in a single ffmpeg invocation, but you have to be clever about writing your filters :)
[22:48:08 CEST] <teratorn> hmm ok, well to finish my idea... I can generate an intermediate stream with my stolen frame for the duration of the pause period, overlay my static-frame-generated-stream on top of it.
[22:48:14 CEST] <teratorn> if that makes any sense at all... :)
[22:52:34 CEST] <teratorn> well, in my use-case, the videos are already separate files... ie. the "pause and overlay" happens at the end of one video, for some number of seconds... then they are all cat'ed together :)
[00:00:00 CEST] --- Wed Oct  5 2016