[Ffmpeg-devel-irc] ffmpeg.log.20200125

[burek]

[00:19:17 CET] <emr> Hello i'm trying to use hw accel, however i got different results here is the story https://privatebin.net/?fb7414da57fedadd#9aUe5Q2UBCgTfX4RJzeGuw8dAmn6tQS9HQreQMGreTMw
[00:22:44 CET] <BtbN> The md5 sum of an mp4 file is not very useful as comparison
[00:23:25 CET] <BtbN> Also, you aren't seriously expecting two lossy encoders to produce bitexactly the same result?
[00:24:57 CET] <emr> but the sizes are also different like  test.mp4 -> 258M and test2.mp4 -> 185M
[00:25:07 CET] <emr> so i was little worried about it
[00:25:39 CET] <BtbN> It's two different encoders after all
[00:25:55 CET] <BtbN> I'm not sure if either of them even care about -qp
[00:26:37 CET] <emr> BtbN, but its still same for viewers right, can i use safely
[00:26:38 CET] <cehoyos> For x264 (and x265), you cannot expect to get identical output on two runs for identical input (if you use multi-threading)
[00:26:59 CET] <BtbN> You will have to look at it to find out how well each encoder did
[00:27:29 CET] <BtbN> it's a lossy codec, quality of encoders varies wildly, specially hardware encoders are pretty much always worse than libx264
[00:27:59 CET] <emr> i got it, thanks BtbN, going to check it
[05:23:54 CET] <Media_Thor>  Greetings! Is there any filter/tools to help identify channels for 6X mono >> 5.1 please? https://trac.ffmpeg.org/wiki/AudioChannelManipulation#a6mono5.1
[07:44:05 CET] <JustLandedOnMars> hi
[07:47:23 CET] <JustLandedOnMars> so I used this command:
[07:47:27 CET] <JustLandedOnMars> ffprobe -loglevel quiet -print_format flat -show_streams file.mp4
[07:47:40 CET] <JustLandedOnMars> and it prints:
[07:47:41 CET] <JustLandedOnMars> streams.stream.0.r_frame_rate="353/12"
[07:48:16 CET] <JustLandedOnMars> so is "353" frame rate correct ?
[07:48:22 CET] <JustLandedOnMars> because if I just use:
[07:48:26 CET] <JustLandedOnMars> ffprobe file.mp4
[07:48:38 CET] <JustLandedOnMars> it prints:
[07:48:40 CET] <JustLandedOnMars> Stream #0:0(und): Video: h264 (Main) (avc1 / 0x31637661), yuv420p(tv, bt709), 1280x720 [SAR 1:1 DAR 16:9], 689 kb/s, 29.42 fps, 29.42 tbr, 90k tbn, 58.82 tbc (default)
[07:48:48 CET] <JustLandedOnMars> here it is: 29.42 fps
[07:48:57 CET] <JustLandedOnMars> so where those 353 came from ?
[07:49:56 CET] <JustLandedOnMars> I tryed some other videos, and it prints:
[07:50:04 CET] <JustLandedOnMars> streams.stream.0.r_frame_rate="2997/100"
[07:50:05 CET] <JustLandedOnMars> streams.stream.0.r_frame_rate="30000/1001"
[07:50:19 CET] <JustLandedOnMars> so how to get actual numbers like 30 fps ?
[08:37:21 CET] <emr> Hello, https://bpaste.net/4BLA i'm losing video when try to encode via h264_vaapi
[11:29:39 CET] <Aelius> is there some disadvantage to the +faststart flag?
[11:30:52 CET] <Aelius> Or is it perhaps just unnecessary? I did not know about it until now, and firefox definitely plays my converted videos before downloading all.
[13:36:38 CET] <DHE> Aelius: you'll want it for streaming, but it requires a rebuild of the file once the job finishes which could take a while for huge files
[13:44:33 CET] <another> JustLandedOnMars: 353/12 = 29.4166666
[13:45:53 CET] <another> Aelius: streaming might work without it if the client uses range-requests
[14:06:05 CET] <mss> i am using the alsa hw:0,0 on my machine to record audio but can't have the audio play on anything except for mpd, anyone can help me with that?
[14:07:09 CET] <mss> i tried aac, mp3 and wav codecs to test, the file size is far more larger than 0k but i get no audio on mpv, mpg123 or the browser.
[14:08:05 CET] <mss> and i can clearly hear the audio recording if i add it to my mpd playlist, but inspecting the file through mpd clients, they say that the audio bitrate is 0khz even though i can clearly hear it.
[14:34:26 CET] <another> meh
[14:34:53 CET] <another> mss: please post your full commandline + ouput to some pastebin site
[14:45:47 CET] <mss> another: `ffmpeg -f x11grab -r 25 -s 1280x800 -i :0 -f alsa -i hw:0,0 test.mkv` this is both audio+video recording, i used just `ffmpeg -f alsa -i hw:0,0 test.wav` for audio only.
[14:45:56 CET] <mss> will upload sample in a second.
[14:52:28 CET] <mss> another: http://0x0.st/iry7.mkv, here's a demonstartion. if you play that through mpv, it won't output any audio but adding it to mpd and playing it will give you audio.
[15:02:33 CET] <debianuser> mss: `mplayer iry7.mkv` plays the audio just fine for me... I can hear someone (you?) clicking on the keyboard and some noise next to a microphone...
[15:03:26 CET] <mss> debianuser: i'm probably having some issues with alsa itself i guess.
[15:03:34 CET] <mss> something is grabbing and not letting alsa go.
[15:04:15 CET] <mss> i'll deal with it with the folks over alsa, i'll come back if i can't figure it out.
[15:04:19 CET] <another> sound works
[15:05:01 CET] <another> what's surprising me, is that ffmpeg chooses ac3
[16:21:42 CET] <cehoyos> another: If you didn't compile FFmpeg with --enable-libvorbis, ac3 is the default for mkv output (the native vorbis encoder is too bad)
[16:22:22 CET] <furq> is there some reason it's not aac other than nobody got around to changing it
[16:27:44 CET] <another> huh
[16:28:32 CET] <another> goes i never had an ffmpeg w/o libvorbis
[17:31:46 CET] <cehoyos> furq: For multi-channel, I believe the ac3 encoder is superior but that doesn't mean you are wrong:-)
[17:32:16 CET] <cehoyos> another: I have to compile it with libvorbis when I see a related bug report, I normally never do it
[17:43:06 CET] <furq> is libvorbis superior for multichannel
[17:44:11 CET] <pink_mist> I thought he just literally said ac3 was superior for multichannel
[17:45:44 CET] <Hello71> well if you're going to change the default, why not just change it to opus (notwithstanding that that's not native either)
[17:47:48 CET] <furq> pink_mist: superior to aac
[17:59:46 CET] <relaxed> Hello71: aac is a native codec and better supported
[18:48:52 CET] <cehoyos> furq: I don't know, I thought the ac3 encoder produces good quality
[22:53:54 CET] <LFSVeteran> modified transcode_aac.c to use PCM_S32LE: https://pastebin.com/q3TZKaCD
[22:54:00 CET] <LFSVeteran> now I get "Could not allocate output frame samples (error 'Invalid argument')"
[22:57:03 CET] <BtbN> Why do you want an encoder for a raw format?
[22:58:49 CET] <LFSVeteran> Maybe some code can be removed but I'm not familiair with it
[22:59:25 CET] <LFSVeteran> what it has to do is "ffmpeg -f s32le -ac 2 -ar <samplerate> -i http://ip:port/stream -acodec pcm_s32le output.wav" in code
[23:00:06 CET] <BtbN> And what part of it fails?
[23:01:08 CET] <LFSVeteran> it seems that it fails at init_output_frame
[23:01:19 CET] <LFSVeteran> row 619
[23:03:28 CET] <BtbN> That part of the code is even fully independend of the "codec" in use
[23:05:14 CET] <LFSVeteran> what I have changed: https://pastebin.com/3mSn7B7J
[23:05:16 CET] <BtbN> I'd print all the parameters it sets and see if one of them is nonsensical
[23:05:56 CET] <LFSVeteran> with AAC it looked like it worked
[23:06:16 CET] <BtbN> so one of the few parameters must be wrong
[23:06:19 CET] <BtbN> easy to check
[23:16:59 CET] <LFSVeteran> nb_samples = 0?
[23:21:54 CET] <CustosLimen> hi
[23:22:06 CET] <CustosLimen> how do I concatenate many mp3s and add silence between them?
[23:27:27 CET] <BtbN> create an mp3 with silence, and put it in between
[23:28:03 CET] <BtbN> LFSVeteran, sounds about wrong
[23:29:52 CET] <DHE> CustosLimen: you probably want to concatenate silence mp3s into the list
[23:30:13 CET] <CustosLimen> ah
[23:30:14 CET] <CustosLimen> good idea
[23:30:28 CET] <LFSVeteran>     (*frame)->nb_samples     = frame_size;
[23:30:28 CET] <LFSVeteran>     (*frame)->channel_layout = output_codec_context->channel_layout;
[23:30:28 CET] <LFSVeteran>     (*frame)->format         = output_codec_context->sample_fmt;
[23:30:28 CET] <LFSVeteran>     (*frame)->sample_rate    = output_codec_context->sample_rate;
[23:30:32 CET] <CustosLimen> now to make silence mp4
[23:30:35 CET] <CustosLimen> mp3*
[23:30:44 CET] <LFSVeteran> are the only one set, and only nb_samples is zero
[23:30:52 CET] <LFSVeteran> guess I have to check that part
[23:31:52 CET] <CustosLimen> got it
[23:33:43 CET] <BtbN> LFSVeteran, well, you are thus asking it to allocate memory for 0 samples. Which makes no sense. Hence the error.
[23:33:52 CET] <LFSVeteran> y
[23:34:07 CET] <LFSVeteran> alloc (0) doesn't make sense
[00:00:00 CET] --- Sun Jan 26 2020