[Ffmpeg-devel-irc] ffmpeg.log.20181107

[burek]

[00:26:37 CET] <benbro> furq: padding/scaling before format seems to fix it
[00:26:39 CET] <benbro> thank you
[00:27:02 CET] <benbro> I still don't understand why it sometimes succeeded even with odd width/height
[03:36:34 CET] <ZeroWalker> can i trim a video so that the audio starts at the same time as the video (currently it's negative, so audio plays a bit before the video, so i just want to cut out that till the video starts)
[03:36:55 CET] <ZeroWalker> so basically, cut audio till the first key frame
[03:52:21 CET] <mixfix41> which video editor would you use just ffmpeg?
[03:52:31 CET] <mixfix41> maybe use a gui video editor
[03:53:06 CET] <frendShot> if he's asking in #ffmpeg, he's probably asking how to use ffmpeg to do that...
[03:56:14 CET] <mixfix41> oh ok so just trim the file to the file starts both audio and video with ffmpeg
[03:57:51 CET] <mixfix41> yea i do sound a bit brash but i was actually visualizing using kdenlive and all but yeah a quick trim off the beginning should do the work
[04:40:26 CET] <mixfix41> ZeroWalker: i think you could use this https://superuser.com/questions/138331/using-ffmpeg-to-cut-up-video for single clips then but later used this https://superuser.com/questions/820747/slicing-video-file-into-several-segments due to a better file/multiple trims
[11:15:16 CET] <mmu_man> plop
[11:15:19 CET] <mmu_man> is trac down?
[11:25:30 CET] <c_14> works here
[11:41:25 CET] <BtbN> It's been insanely slow the last couple days for me
[11:42:08 CET] <zerodefect> In what respect?
[11:44:36 CET] <zerodefect> Using the C-API and when dealing with sRGB, what do I set the AVFrame's color primaries too (ie. AVColorPrimaries). AVCOL_PRI_BT709 ?
[13:03:56 CET] <zerodefect> Hmmm...and if want to convert between bt.709 and bt.601, can I get away with swscale or am I forced to utilize colorspace filter.
[13:05:17 CET] <furq> zerodefect: use colorspace or zscale
[13:07:02 CET] <zerodefect> Does swscale not make the grade?
[13:07:59 CET] <zerodefect> My hesitation with using filters is that the graphs are very static....cannot change input parameters without recreating graph.
[13:08:44 CET] <furq> you could also just flag the output as 709 if that's an option
[13:09:28 CET] <zerodefect> Not sure what you mean, can you elaborate?
[13:10:56 CET] <furq> i mean just not converting colourspace and setting the output primaries etc to 709
[13:11:11 CET] <furq> if that's not an option then you're stuck with using filters afaik
[13:12:17 CET] <zerodefect> Ok. I guess I ask this because I note that swscale has a `sws_setColorspaceDetails(..)` function.
[13:17:29 CET] <mmu_man> [ogg @ 0x5597871796c0] Error reading attached picture data. 708 != 5120
[13:17:44 CET] <mmu_man> no wonder I can't get a thumbnail inside
[13:17:49 CET] <mmu_man> seems to be truncated
[13:32:44 CET] <mmu_man>     uint8_t line[1024];
[13:32:46 CET] <mmu_man> bleh&
[13:34:33 CET] <mmu_man> oh
[13:35:35 CET] <mmu_man> &
[13:38:22 CET] <mmu_man> \o/
[13:38:30 CET] <mmu_man> en forçant le buffer à 256k ça passe
[13:38:40 CET] <mmu_man> sorry
[13:38:49 CET] <mmu_man> forcing the buffer size to 256k worked
[13:44:42 CET] <mmu_man> hmm aviobuff:ff_get_line & ffmetadec:get_line could use some factoring
[14:51:05 CET] <mmu_man> patch sent
[14:54:28 CET] <wr> hello, need a help on converting a mp4 video file on the ffmpeg to ogv, how can do it?
[15:29:52 CET] <saml> wr, ffmpeg -i a.mp4 -c:v libtheora -c:a libvorbis a.ogv
[15:30:37 CET] <wr> ah thanks
[15:54:50 CET] <Fenrirthviti> Hey folks, I know this isn't really your problem, but looking for any leads on what might be causing this crash: https://obsproject.com/logs/9OJ_4cLDPVgO6i8h
[15:55:02 CET] <Fenrirthviti> NVENC startup crash when using ffmpeg as the output
[15:55:51 CET] <Fenrirthviti> Curious if anything similar has been seen with ffmpeg natively or not
[16:20:31 CET] <BtbN> Impossible to say without debug symbols, but I'd blame a broken driver.
[16:20:48 CET] <BtbN> Or really horribly broken parameters.
[16:23:33 CET] <zerodefect> Agreed about debug symbols. c0000005 means `Access Violation` from memory. It's not very happy.
[16:24:54 CET] <Mavrik> Feels violated.
[16:32:03 CET] <wr> saml, the video just turned to be more low quality than the mp4, any way i increase it?
[16:32:46 CET] <relaxed> wr: https://trac.ffmpeg.org/wiki/TheoraVorbisEncodingGuide
[16:33:17 CET] <saml> you mean it became retro and better
[16:33:35 CET] <wr> saml, too much retro lol
[16:45:49 CET] <kepstin> well, you asked for theora, you got retro ;)
[16:47:37 CET] <kepstin> why do you want theora anyways? pretty much the only thing that's needed for is obsolete versions of firefox, afaik.
[18:49:56 CET] <pi-> I'm experiencing a strange problem.
[18:50:07 CET] <pi-> I have a source .caf waveform.
[18:50:42 CET] <pi-> If I convert it to .wav with Audacity (which if I remember correctly uses ffmpeg), my DSP code processes it correctly.
[18:51:09 CET] <pi-> However if I use `ffmpeg -i X.caf X.wav`, the code now fails to process it correctly.
[18:52:07 CET] <kepstin> can you pastebin the output of that ffmpeg command please?
[18:52:38 CET] <kepstin> and also what do you mean by "fails to process it correctly?" how/where does it fail?
[18:54:43 CET] <kepstin> with that command, ffmpeg will convert the audio to 16-bit little-endian pcm in the wav file, but it should preserve the sample rate and number of channels from the input.
[18:56:41 CET] <pi-> kepstin: https://paste.pound-python.org/show/44sRGMU0wu3bZ3eRysnL/
[18:57:43 CET] <pi-> If I examine hexdumps of the generated WAVs (both from audacity and from ffmpeg), they do look strange.  Every single 4-cluster ends ??00 or ??ff.
[18:58:10 CET] <kepstin> yeah, that's converting from big-endian audio (in the caf) to little-endian (in the wav). If your DSP code expects the opposite endian and doesn't read the wav headers properly that could cause issues
[18:58:38 CET] <pi-> I am attempting to extract an ultrasound signal, so 'fails' is difficult to define.
[18:58:51 CET] <pi-> It is actually a partial failure.
[18:59:05 CET] <pi-> One part of the signal gets recognised successfully, another part doesn't.
[18:59:42 CET] <pi-> I wrote my own WAV reader, so it's possible I'm not processing the header right.
[18:59:50 CET] <kepstin> how in the world would you get ultrasound stuff from a 48kHz audio file? if it's lowpassed correctly, it won't have anything above 24kHz.
[19:00:00 CET] <pi-> well near-ultra
[19:00:03 CET] <pi-> 18-20kHz
[19:00:11 CET] <kepstin> ah, yeah.
[19:00:40 CET] <kepstin> one thing you can try is to add "-c:a pcm_s16be" to the ffmpeg command, to leave the audio in big-endian in the wav file, see if that helps
[19:01:23 CET] <kepstin> if it does, then your app is expecting big-endian audio, and it isn't reading the wav headers correctly or isn't doing the required conversion before processing.
[19:02:20 CET] <kepstin> oh, nvm, wav doesn't support big endian at all
[19:02:23 CET] <kepstin> my mistake
[19:02:27 CET] <pi-> the endianness appears to be consistent between ffmpeg and audacity output.
[19:02:30 CET] <kepstin> (it's a PC/windows thing)
[19:03:05 CET] <pi-> I'm guessing the xx00 and xxff everywhere means small positive and small negative numbers
[19:03:37 CET] <pi-> i.e. between -/+ 255 over a total range of -/+ 32768
[19:18:19 CET] <kepstin> yeah, that makes sense.
[19:19:50 CET] <kepstin> in the caf file, i'd expect the values to all look like 00xx or FFxx
[19:30:39 CET] <GlidingHorse> hi, can I mux HEVC into an mpegts with ffmpeg 4.0?
[19:31:03 CET] <GlidingHorse> i'm not asking how ,just asking if it's supported
[19:31:13 CET] <durandal_1707> no
[19:31:45 CET] <GlidingHorse> :(
[19:31:59 CET] <kepstin> pi-: hmm. one thing to note is that ffmpeg usually adds some metadata to the wav file, you want to make sure your wav reader correctly locates the start of the "data" chunk.
[19:32:32 CET] <durandal_1707> GlidingHorse: or yes, dunno i thought you ask about av1
[19:32:55 CET] <GlidingHorse> :)
[19:33:47 CET] <GlidingHorse> anyway i will try by myself, but wanted to gain some time while ffmpeg downloads
[19:34:15 CET] <durandal_1707> it should be possible, i see lines with HEVC in mpegts muxer
[19:34:40 CET] <GlidingHorse> excelent :D
[20:08:25 CET] <GlidingHorse> indeed, ffmpeg muxed hevc video and e-ac-3 audio into mpegts :D
[20:08:35 CET] <GlidingHorse> thanks for the support :D
[20:08:37 CET] <GlidingHorse> o/
[20:08:38 CET] <GlidingHorse> \part
[00:00:00 CET] --- Thu Nov  8 2018