[Ffmpeg-devel-irc] ffmpeg.log.20160807

burek burek021 at gmail.com
Mon Aug 8 03:05:01 EEST 2016


[02:31:25 CEST] <ajf> I want to convert a video file from h.264 to MJPEG, just for the heck of it. How do I tell ffmpeg to maintain the source's bitrate?
[02:32:05 CEST] <c_14> you can't
[02:32:36 CEST] <c_14> you could technically check the size of the input file, do math and get the bitrate and tell ffmpeg to use that
[02:32:39 CEST] <c_14> But it won't be the same quality
[02:32:46 CEST] <ajf> oh sure
[02:32:54 CEST] <c_14> And unless the input is constant bitrate, it won't match anyway
[02:32:58 CEST] <ajf> right
[02:33:26 CEST] <c_14> If this is just random testing whatever, check the output of ffprobe if it gives you a bitrate on the input, then subtract the audio stream and round down a bit
[02:34:15 CEST] <ajf> hmm
[02:34:26 CEST] <ajf> does -maxbitrate apply to the entire stream, or just video?
[02:35:45 CEST] <c_14> Do you mean -maxrate ?
[02:35:48 CEST] <c_14> There's no -maxbitrate
[02:35:51 CEST] <ajf> probably
[02:36:18 CEST] <c_14> I think it only applies to the video stream, I'm not sure though.
[02:36:40 CEST] <ajf> oh wait I think I want -b:v
[02:36:46 CEST] <ajf> and that sets video bitrate specifically
[02:38:09 CEST] <c_14> yes
[02:40:36 CEST] <c_14> Though I'm not sure the mjpeg encoder has a cbr mode...
[02:41:26 CEST] <ajf> hah, yep
[02:41:35 CEST] <ajf> I think it's completely ignored me
[02:41:37 CEST] <ajf> :D
[02:42:34 CEST] <c_14> you might be able to use -q or -qp to set a quantizer and play around with that until it's about the same size
[02:46:49 CEST] <ajf> yeah. it looks like the mjpeg encoder really doesn't like being told what bitrate to use
[02:47:04 CEST] <ajf> which does make sense in fairness
[02:52:16 CEST] <ajf> huh, mjpeg doesn't like -qp
[02:54:12 CEST] <c_14> -q seems to work
[03:02:09 CEST] <ajf> it looks like mjpeg is really stubborn
[03:02:10 CEST] <ajf> :<
[03:02:55 CEST] <ajf> I mean I'd happily tell it to encode everything as JPEGs of quality 10 (extreme quantisation)
[03:02:59 CEST] <ajf> but there's no such option
[03:04:32 CEST] <c_14> set -q:v to 31
[03:04:43 CEST] <c_14> I think that's the lower end of the range
[03:04:52 CEST] <ajf> yeah
[03:04:56 CEST] <ajf> it is also the default
[03:27:33 CEST] <ajf> thanks for your attempts to help me
[05:03:56 CEST] <Spring> is there a bulletproof syntax for filter_complex commands?
[05:04:02 CEST] <Spring> using the command line
[05:04:47 CEST] <Spring> I've escaped both colons and commas (not that they even need to be escaped but I've tried everything) and keep getting errors
[05:05:39 CEST] <Spring> the comment in parenthesis above referring to the commas that is
[05:10:16 CEST] <durandal_1707> using script files
[05:10:50 CEST] <Spring> durandal_1707, I'm using batch
[05:11:27 CEST] <Spring> ffmpeg keeps detecting the entire line as the arguments for a single command
[05:11:49 CEST] <durandal_1707> I mean ffmpeg option
[05:12:55 CEST] <Spring> is there a doc page on this, nothing is showing up in a search
[05:16:30 CEST] <durandal_1707> -filter-complex-script?
[05:20:15 CEST] <Spring> this? https://trac.ffmpeg.org/wiki/FilteringGuide#Scriptingyourcommandlineparameters
[05:20:19 CEST] <Spring> can't find anything else
[05:21:12 CEST] <wallbroken> is possible to convert from pcm to aac ?
[05:21:18 CEST] <durandal_1707> its not in wiki, :(
[05:22:06 CEST] <durandal_1707> Spring: search ffmpeg docs on ffmpeg.org page
[05:22:39 CEST] <durandal_1707> wallbroken: yes, its possible
[05:22:48 CEST] <wallbroken> how?
[05:22:59 CEST] <Spring> woah, literally has a single line about it mentioned nowhere else
[05:23:10 CEST] <Spring> -filter_script
[05:23:32 CEST] <Spring> er, rather, -filter_complex_script
[05:24:11 CEST] <durandal_1707> wallbroken: add -c:a aac
[05:25:28 CEST] <wallbroken> durandal_1707, it's a little harder, the PCM stream is in an .avi file
[05:26:27 CEST] <durandal_1707> wallbroken: what container?
[05:26:38 CEST] <wallbroken> .avi
[05:26:47 CEST] <durandal_1707> for output?
[05:27:21 CEST] <wallbroken> for output must be .aac format
[05:27:29 CEST] <wallbroken> the input is a pcm in a .avi file
[05:28:28 CEST] <durandal_1707> than just: ffmpeg -i in.AVI out.aac for start
[05:29:09 CEST] <wallbroken> https://www.dropbox.com/s/5evj7aijndgvzt6/ok.txt?dl=0
[05:29:13 CEST] <wallbroken> this is the input file
[05:29:23 CEST] <wallbroken> can you tell me the entire command?
[05:29:27 CEST] <furq> he just did
[05:29:47 CEST] <wallbroken> but i need to tell the bitrate
[05:29:58 CEST] <furq> -b:a 128k
[05:30:21 CEST] <Spring> alternative question, can the trim filter be configured to cut a certain part of a video according to the original timestamp, then that trimmed portion speed adjusted in a single command?
[05:30:36 CEST] <Spring> *single filtergraph line
[05:30:44 CEST] <wallbroken> " ffmpeg -i in.AVI -b:a 128k out.aac" that's all?
[05:31:20 CEST] <furq> yes
[05:31:39 CEST] <Spring> even using seconds instead of the full timestamp (which even escaped gives errors) doesn't trim the video correctly before applying the speed filter
[05:32:01 CEST] <wallbroken> furq, PCM it's lossy?
[05:33:19 CEST] <durandal_1707> Spring: you do not escape them.when using script option
[05:34:14 CEST] <Spring> I kind of need the batch script to be one file though, and as mentioned even using seconds doesn't produce the right output (eg: using '64' instead of 1:04)
[05:35:27 CEST] <Spring> eg http://pastebin.com/7ukXsxMd
[05:36:01 CEST] <Spring> this produces a 3:48 video clip of a static frame
[05:37:20 CEST] <Spring> or a /super/ slow video it's hard to tell
[05:37:34 CEST] <durandal_1707> perhaps you give args for exact frame number
[05:38:06 CEST] <furq> don't you need atrim as well as trim
[05:38:36 CEST] <Spring> yes but since trim was causing trouble I wanted to figure out one at a time
[05:39:27 CEST] <Spring> I also need timestamps rather than frame number, since that's what is used for the input initially
[05:41:09 CEST] <Spring> I guess I'm just completely puzzled by the trim filter, since the others work fine without it, whereas '-ss blah to blah' works but is affected by the speed filter so it outputs the wrong portion
[05:42:04 CEST] <Spring> hence why I needed to see if the trim filter could be forced to trim the video before applying the speed filter
[05:42:29 CEST] <durandal_1707> Spring: what ffmpeg version?
[05:43:02 CEST] <durandal_1707> perhaps you need to reset video pts to 0
[05:43:22 CEST] <Spring> durandal_1707, version N-80801-gc0cb53c
[05:43:32 CEST] <durandal_1707> not just speed it up
[05:44:50 CEST] <durandal_1707> because ffmpeg will try to keep audio and video in sync
[05:45:29 CEST] <durandal_1707> and trim doesn't reset pts
[05:48:28 CEST] <furq> Spring: -vf trim=30:40,setpts=2*(PTS-STARTPTS) -af atrim=30:40,atempo=0.500
[05:48:29 CEST] <furq> that works for me
[05:49:48 CEST] <durandal_1707> Yeah
[05:49:57 CEST] <Spring> I'm actually wondering if it's interpreting 2.000 as '2000' not '2.0'
[05:50:28 CEST] <furq> it isn't
[05:50:34 CEST] <furq> you need to subtract STARTPTS or it won't work
[05:50:45 CEST] <furq> or reset the pts in some other way, but that seems easiest
[05:52:13 CEST] <furq> the command you pasted will work but the video will start playing at 2:08
[05:53:01 CEST] <wallbroken> durandal_1707, it does not work
[05:53:09 CEST] <wallbroken> the .avi file has two streams in
[05:53:43 CEST] <wallbroken> please take a look at those mediainfo input file: https://www.dropbox.com/s/5evj7aijndgvzt6/ok.txt?dl=0
[05:54:08 CEST] <wallbroken> the play of the output file does not produce anything
[05:54:28 CEST] <wallbroken> the file lenght is 50 mins, but it does nor reproduce anything
[05:54:41 CEST] <durandal_1707> wallbroken: what's output of ffmpeg command?
[05:55:44 CEST] <durandal_1707> mediainfo is useless to me, better use ffprobe....
[05:56:02 CEST] <wallbroken> https://www.dropbox.com/s/py68bqqu8bnp5qa/out.txt?dl=0
[05:57:06 CEST] <Spring> furq, you don't know how much you relieved me with that answer, thanks a TON
[05:57:17 CEST] <furq> Spring: -vf trim=00\\:00\\:30:00\\:00\\:40
[05:57:20 CEST] <furq> that also works for me fwiw
[05:57:58 CEST] <furq> although it's probably easier to just convert to seconds rather than figure out what escaping level you're at
[05:58:10 CEST] <furq> it is nice to know that it technically does accept duration specs though
[05:59:01 CEST] <Spring> hmm, may try testing the escaped shorthand
[05:59:09 CEST] <wallbroken> durandal_1707, you saw it?
[05:59:23 CEST] <wallbroken> *seen
[05:59:37 CEST] <furq> wallbroken: that looks like it's working fine
[05:59:40 CEST] <furq> what exactly is the problem
[05:59:48 CEST] <durandal_1707> yes, what you use for playback?
[06:00:20 CEST] <wallbroken> furq, open it with vlc
[06:00:36 CEST] <wallbroken> play, the playback is about 4 seconds, with no sound
[06:00:39 CEST] <wallbroken> then stops it
[06:00:52 CEST] <wallbroken> but the duration of all the track is about 50 minutes
[06:00:58 CEST] <wallbroken> and in vlc is also reported
[06:01:37 CEST] <Spring> does the shorthand trim=timecode interpret a single timecode as inherently representing the start (without an end trim)?
[06:02:09 CEST] <wallbroken> furq, wait, now it seem to work but the duration is continuosly changing
[06:02:10 CEST] <furq> The defaults are such that all the input is kept. So it is possible to set e.g. just the end values to keep everything before the specified time.
[06:03:24 CEST] <furq> wallbroken: have you tried playing it in something better than vlc
[06:03:51 CEST] <wallbroken> furq, vlc is the masterpice of playback software
[06:03:58 CEST] <wallbroken> *masterpiece
[06:04:38 CEST] <furq> i'll take your word for it
[06:04:53 CEST] <durandal_1707> lol
[06:05:27 CEST] <wallbroken> furq, that's all people said to me: "vlc playpacks also rocks"
[06:06:23 CEST] <furq> there's no accounting for taste
[06:06:45 CEST] <furq> try it in ffplay, mpv, fb2k or something competent
[06:07:17 CEST] <wallbroken> i don't know, but which software should be better? windows media player? quicktime player?
[06:07:31 CEST] <furq> mpv or mpc-hc on windows
[06:09:53 CEST] <wallbroken> somebody can tell me once for all, if is preferred ffmpeg 32bit or 64?
[06:10:48 CEST] <durandal_1707> 64 on 64 bit, 32 bit otherwise
[06:11:17 CEST] <wallbroken> 32 bit encoded video could be not compatible with some platform?
[06:11:18 CEST] <Spring> I use PotPlayer
[06:11:34 CEST] <furq> wallbroken: no
[06:11:35 CEST] <furq> that would be insane
[06:11:46 CEST] <wallbroken> what?
[06:12:37 CEST] <durandal_1707> Isn't potplayer license violator?
[06:13:19 CEST] <wallbroken> another thing: i want to convert a vob video in x264 with quality factor: 23, how to do it?
[06:13:41 CEST] <furq> -i src.vob out.mkv
[06:13:45 CEST] <Spring> not sure, heard something like that. Best player I've used on Windows though.
[06:13:58 CEST] <wallbroken> furq, and about the quality factor
[06:14:00 CEST] <wallbroken> and the preset?
[06:14:12 CEST] <furq> the default crf is 23
[06:14:19 CEST] <wallbroken> and the preset?
[06:14:22 CEST] <furq> medium
[06:14:25 CEST] <wallbroken> if i want to use placebo?
[06:14:29 CEST] <furq> -preset placebo
[06:14:37 CEST] <furq> note that i am not recommending preset placebo
[06:14:54 CEST] <wallbroken> just for the high time needed?
[06:15:08 CEST] <furq> the clue's in the name
[06:15:41 CEST] <furq> if you want to spend a million years encoding then at least use x265 and get some benefit out of it
[06:16:30 CEST] <wallbroken> x265?
[06:16:36 CEST] <wallbroken> is better than x264 ?
[06:16:44 CEST] <furq> it's one better than x264
[06:17:01 CEST] <wallbroken> ok, and how to use it?
[06:17:53 CEST] <furq> https://trac.ffmpeg.org/wiki/Encode/H.265
[06:18:22 CEST] <wallbroken> and between mp4 and mkv
[06:18:27 CEST] <wallbroken> which one is better?
[06:18:57 CEST] <wallbroken> something buggy in aac of ffmpeg
[06:19:05 CEST] <wallbroken> i tried opening the same file with groove music
[06:19:13 CEST] <wallbroken> and the wrong duration time here too
[06:21:23 CEST] <wallbroken> let me try doing it in mp3
[06:48:37 CEST] <flai> Hey guys, I want to encode a video for use in a HTML5-<video /> element. Currently I'm trying to use mp4 with h264/aac. The problem is that the video-file isn't working, and I can't figure out why. Details: (Picture, Commandline, ...): https://gist.github.com/moritzuehling/3a08559340910721079811069768f751
[06:49:12 CEST] <flai> Any help and pointer on why my encoded video isn't working and what I can do to make it work would be very appreciated!
[08:52:56 CEST] <hollabackatcha> can ffsever be used for live streaming ? right now it's just reloading inputed videos from the beginning... it's not live
[09:16:34 CEST] <Spring> can the atempo filter be stacked for slow mo, as well? Using atempo=.500,atempo=.500 gives an error
[09:23:28 CEST] <c_14> Spring: it should work (it works here)
[09:24:40 CEST] <Spring> c_14, did you test using the -filter:a option?
[09:24:54 CEST] <c_14> I used -af (which should be the same thing)
[09:25:11 CEST] <Spring> I'm adding mine within -filter_complex
[09:25:41 CEST] <c_14> Works with a filter_complex as well
[09:28:57 CEST] <Spring> c_14, http://pastebin.com/DsnLTESy
[09:30:05 CEST] <Spring> ffmpeg crashes as soon as the encode begins
[09:32:22 CEST] <c_14> The problem is the atrim filter
[09:32:26 CEST] <c_14> If you remove that it works
[09:32:43 CEST] <Spring> it works if the second atempo isn't there
[09:33:10 CEST] <c_14> not on my system
[09:33:33 CEST] <Spring> actually, you're right, wtf
[09:34:23 CEST] <c_14> eh
[09:34:34 CEST] <furq> that works here
[09:34:38 CEST] <c_14> The atrim filter doesn't accept timestamps in HH:MM:SS.MSS notation
[09:34:46 CEST] <furq> oh
[09:34:58 CEST] <c_14> hmm, no it should
[09:34:58 CEST] <Spring> WUT
[09:35:06 CEST] <furq> -af atrim=0,atempo=0.5,atempo=0.5
[09:35:08 CEST] <c_14> the escaping just isn't working
[09:35:08 CEST] <furq> that works for me
[09:35:39 CEST] <furq> Spring: fwiw if those are the only filters you're using then there's no need to use filter_complex
[09:35:44 CEST] <c_14> adding single quotes around the timestamp in addition to the backslashes works for me
[09:35:50 CEST] <c_14> "[0:a]atrim='00\:00\:00.000',atempo=.5,atempo=.5[a]"
[09:35:52 CEST] <c_14> ^like that
[09:36:59 CEST] <c_14> The colons weren't escaped correctly, so the atrim filter never produced output which caused the assertion in the atempo filter
[09:38:19 CEST] <furq> also it's a waste of time to run that in the first pass
[09:38:28 CEST] <furq> just use -an to ignore the audio
[09:39:06 CEST] <c_14> Ye, for pass 1 you can just ignore the audio since that isn't taken into account for the 2pass data generation
[09:39:24 CEST] <Spring> I had -an originally, can't recall why I removed it, possibly while encountering some other error
[09:40:44 CEST] <Spring> and yup, the single quotes fixed the error, cheers!
[09:41:31 CEST] <Spring> is there a technical reason why atempo can't go below .5 or above 2.0 in one run?
[09:47:31 CEST] <LJHSLDJHSDLJH> I'm trying to reduce video files size using the below command line and it's taking forever to complete and the amount of reduced file is size is very disappointing. Can anyone help me with better syntax maybe!
[09:47:35 CEST] <LJHSLDJHSDLJH> for x in *.mp4;do ffmpeg -i "$x" -c:v libx264 -crf 24 -b:v 1M -c:a aac -strict -2 "./x/$x";done
[09:48:53 CEST] <cbsrobot> LJHSLDJHSDLJH: If possible update the ffmpeg version
[09:49:22 CEST] <cbsrobot> aac encoder got a lot better lately and is not experimental anymore
[09:50:08 CEST] <cbsrobot> so you can drop the " -strict -2"
[09:50:33 CEST] <cbsrobot> for the video drop the " -b:v 1M", just play with the -crf value
[09:51:27 CEST] <cbsrobot> I suggest try values between 27 (high compression, small filesize) and 18 (smaller compression, bigger filesize)
[09:59:50 CEST] <LJHSLDJHSDLJH> hello cbsrobot : ffmpeg version 2.8.6-1ubuntu2 Copyright (c) 2000-2016 the FFmpeg developers
[09:59:57 CEST] <LJHSLDJHSDLJH> do I need to upgrade that?
[10:01:44 CEST] <cbsrobot> LJHSLDJHSDLJH: I do not remember the exact version the aac encoder was updated, but drop the "-strict -2" and if it still works then you should be good to go
[10:02:01 CEST] <furq> it was upgraded in 3.0
[10:02:13 CEST] <furq> but if the source files are mp4 then there's probably no point reencoding the audio
[10:02:24 CEST] <furq> just use -c:a copy
[10:05:09 CEST] <LJHSLDJHSDLJH> alright, now it's working again .. I'll wait and see the results , thanks
[10:05:11 CEST] <LJHSLDJHSDLJH> for y in *.mp4;do ffmpeg -i "$y" -c:v libx264 -crf 20 -c:a copy "./y/$y";done
[10:21:51 CEST] <xochilpili> hi all
[10:22:13 CEST] <xochilpili> im having this trouble, i want to use vod (video on demand) using vlc
[10:22:32 CEST] <xochilpili> i have use: cvlc /media/Movies/Che.el.Argentino.avi --sout '#transcode{vcodec=libx264,vb=2000,scale=0,channels=2,samplerate=44100}:std{access=http{mime=video/mp4},mux=mp4,dst=:8080/cam1}' -vvv
[10:22:42 CEST] <xochilpili> and this is the full output: http://pastebin.com/HEC4GSyj
[10:22:53 CEST] <xochilpili> please any help i will appreciate :D
[10:26:25 CEST] <cbsrobot> xochilpili: maybe ask in #videolan
[10:26:35 CEST] <xochilpili> im in there
[10:26:38 CEST] <xochilpili> nothing...
[10:26:59 CEST] <xochilpili> " I also dont know what your problem could be ;) "
[10:27:04 CEST] <xochilpili> they said
[11:50:28 CEST] <LJHSLDJHSDLJH> using the below is too slow as well and moreover it exceeded the original file size
[11:50:31 CEST] <LJHSLDJHSDLJH> for y in *.mp4;do ffmpeg -i "$y" -c:v libx264 -crf 20 -c:a copy "./y/$y";done
[11:51:27 CEST] <LJHSLDJHSDLJH> using this is quickly BUT it converted 762mb file into 2.4gb file size
[11:51:28 CEST] <LJHSLDJHSDLJH> for z in *.mp4;do ffmpeg -i "$z" -c:v libx264 -preset ultrafast -crf 20 -c:a copy "./z/$z";done
[11:52:03 CEST] <LJHSLDJHSDLJH> therefore ffmpeg conversion is considered crap at my side until further change happen
[11:52:19 CEST] <LJHSLDJHSDLJH> I'll have to redownload the whole thing again
[12:56:01 CEST] <flai> Hey guys, I want to encode a video for use in a HTML5-<video /> element. Currently I'm trying to use mp4 with h264/aac. The problem is that the video-file isn't working, and I can't figure out why. Details: (Picture, Commandline, ...): https://gist.github.com/moritzuehling/3a08559340910721079811069768f751
[12:56:03 CEST] <flai> Any help and pointer on why my encoded video isn't working and what I can do to make it work would be very appreciated!
[12:56:14 CEST] <flai> (I hope it was okay to bump after 6 hours)
[13:21:15 CEST] <BtbN> Both files work perfectly fine for me.
[13:21:46 CEST] <BtbN> Whatever Browser that is, I'd blame it for refusing to play the second file.
[13:24:48 CEST] <furq> flai: pastebin the output of ffprobe -show_streams and/or mediainfo
[13:37:54 CEST] <Spring> is there a calculation for determining atempo values larger than 2.0?
[13:38:29 CEST] <jookiyaya> big news: zen destroys intel  http://static2.gamespot.com/uploads/scale_super/460/4601080/3046688-8311357204-AMD-Z.jpg
[13:41:23 CEST] <flai> furq: https://gist.github.com/moritzuehling/e105d2dd67402bbcf12f15eecdc6897b
[13:42:42 CEST] <furq> maybe try without -level 3.1
[13:42:54 CEST] <furq> i would have assumed ffmpeg would just ignore it for 720p60 but evidently not
[13:47:03 CEST] <flai> furq: Encoding...
[13:54:50 CEST] <flai> furq: Not working :( https://gist.github.com/moritzuehling/60e83cdfbec40a10db33a5a8b5f9e531
[13:55:00 CEST] <jkqxz> jookiyaya:  Zen adds the hardware instructions for hashing functions, so obviously it can do them at close to the memory line rate.  (Skylake and Goldmont have them too - presumably that benchmark is deliberately not using them on the 6700K to make the results more amusingly misleading.)
[13:57:58 CEST] <BtbN> I also have a hard time figuring out what is even compared there. Some MB/s value. of what? Seems rigged to me. Specially as the AMD machine has the fastest RAM of all of them installed.
[13:58:00 CEST] <flai> Maybe it helps: The source-file is recorded with OBS + NVENC...
[13:58:42 CEST] <furq> BtbN: it's SHA1 hashing speed
[13:59:01 CEST] <furq> so yeah, pretty much worthless
[13:59:22 CEST] <BtbN> Then it's even more stupid to give the AMD machine the fastest RAM
[14:00:52 CEST] <jookiyaya> #5 on that list has ddr4 too
[14:00:58 CEST] <furq> it's actually pretty smart to do that because it makes it easier to produce a screenshot where it's at the top of the list
[14:02:06 CEST] <furq> jookiyaya: the ram speed doesn't really matter, sha1 hashing speed is a worthless metric
[14:02:19 CEST] <furq> especially when some of those cpus have extensions for it and some don't
[14:02:54 CEST] <furq> i have to suspect i'm being overly charitable by even explaining this
[14:03:29 CEST] <jookiyaya> then basically, any benchmark score is useless info to you
[14:03:38 CEST] <jookiyaya> since it's only one aspect
[14:03:49 CEST] <BtbN> The score of a pointless benchmark is indeed useless, yes.
[14:04:24 CEST] <BtbN> No CPU that would be a "worthy opponent" in that benchmark is in the list.
[14:04:33 CEST] <jkqxz> In this channel, we would probably be more interested in a benchmark of x264 performance...
[14:04:45 CEST] <BtbN> Like, a i7-6700k or i7-6900k
[14:05:30 CEST] <BtbN> Wait, there is a 6700k.
[14:05:35 CEST] <jkqxz> There's a 6700K there.   Presumably they avoided using the hardware instructions on it in order to ensure the right horse won (it does look like it's in roughly the right place for a software implementation).
[14:05:38 CEST] <BtbN> Yeah, it seems indeed rigged then.
[14:05:48 CEST] <jookiyaya> jkqxz why still use x264, when there is x265
[14:06:31 CEST] <Spring> 1.7 too fast 1.5 too slow, 1.6 too slow... :/
[14:06:54 CEST] <jookiyaya> spring too fast?
[14:07:07 CEST] <jkqxz> x264 has better tuning to occupy the processor as much as possible, so it will be a better benchmark.  Also 9001 times as many people use it, which probably makes it more relevant...
[14:07:14 CEST] <flai> furq: Any idea where the problem may lay?
[14:07:20 CEST] <furq> nope
[14:07:32 CEST] <furq> well
[14:07:34 CEST] <jookiyaya> spring 1.7 is not too fast
[14:07:39 CEST] <furq> if i had to guess i'd say edge is the problem
[14:07:43 CEST] <furq> but you can't really do anything about that
[14:08:00 CEST] <Spring> jookiyaya, can't find the correct calculation for atempo when stacking values over 2.0 (ie: 200%)
[14:08:02 CEST] <jookiyaya> spring is that a typo
[14:08:19 CEST] <furq> Spring: what final value do you want
[14:08:33 CEST] <jookiyaya> spring, huh?  please explain why you said  1.7 = too fast
[14:09:34 CEST] <furq> if you want 3.0 then use 2.0 and 1.5
[14:10:21 CEST] <jookiyaya> furq do you have any idea what spring is saying
[14:10:27 CEST] <furq> yes
[14:10:34 CEST] <Spring> furq, so I'm using an input value (percentage) then calculating the setpts and atempo values from it. However going above 2.0 I can't find the right additional atempo value to add to it
[14:10:43 CEST] <furq> i know
[14:10:50 CEST] <furq> you were perfectly clear
[14:10:57 CEST] <jookiyaya> furq  how is 1.7 too fast, i don't get it
[14:11:25 CEST] <Spring> was just letting off a little steam while looking for 280%
[14:11:38 CEST] <furq> 2.0 and 1.4
[14:14:36 CEST] <Spring> furq, way too slow
[14:15:42 CEST] <jookiyaya> spring so only 1.7 is fast?
[14:16:21 CEST] <jookiyaya> furq  how is 1.7 too fast, i don't get it
[14:16:47 CEST] <Spring> there's some spot between 1.6 and 1.7 that seems to be the correct value for atempo at 280% video speed, but I can't tell what the calculation is meant to be
[14:17:25 CEST] <furq> i don't get how 2.0 and 1.4 doesn't work
[14:18:15 CEST] <Spring> it seems like it would be simple but the results say otherwise
[14:19:36 CEST] <jookiyaya> what are you 2 talking about
[14:21:06 CEST] <Spring> jookiyaya, atempo (audio tempo filter) has a usable range of 0.5 to 2.0. Any lower or higher and you have stack the filters to achieve the desired speed (eg: over 200%/2.0 or under 50%/0.5)
[14:21:55 CEST] <jookiyaya> spring english please,   example "encoding is too slow"   picture looks crappy
[14:22:08 CEST] <jookiyaya> audio and video doesn't sync/match
[14:23:20 CEST] <Spring> basically audio isn't the correct speed for the video, so they don't match
[14:23:45 CEST] <jookiyaya> spring  for which x265 versions?
[14:24:34 CEST] <Spring> it's not about a particular codec
[14:24:35 CEST] <jookiyaya> i've been using x265 since 1.5,  how come i never had this problem
[14:24:43 CEST] <jookiyaya> i've been using x265 since 1.5 to 1.9,  how come i never had this problem
[14:25:08 CEST] <jookiyaya> spring then why did you say this?   <Spring> 1.7 too fast 1.5 too slow, 1.6 too slow
[14:25:39 CEST] <jookiyaya> 1.7 magically became too fast for you?  are you serious
[14:26:04 CEST] <jookiyaya> furq any idea
[14:26:26 CEST] <jookiyaya> dv any idea what spring is talking about
[14:26:39 CEST] <jookiyaya> fahadash any idea what spring is talking about
[14:26:46 CEST] <jookiyaya> durandal_1707 any idea what spring is talking about
[14:26:52 CEST] <jookiyaya> ozette any idea what spring is talking about
[14:27:05 CEST] <Spring> ...I interjected at an inopportune moment, pay my issue no mind
[14:27:17 CEST] <jookiyaya> spring no you said it
[14:27:23 CEST] <jookiyaya> i want my explanation
[14:27:37 CEST] <Spring> I already explained it like twice, what don't you understand?
[14:27:49 CEST] <jookiyaya> you didn't explain in english
[14:28:15 CEST] <jookiyaya> that most people would understand
[14:29:00 CEST] <jookiyaya> <Spring> 1.7 too fast 1.5 too slow, 1.6 too slow ;   1.7 magically became too fast for you?   explain
[14:29:12 CEST] <durandal_1707> it can't match, he is not using interpolation
[14:29:41 CEST] <jookiyaya> durandal_1707  any idea what spring is saying
[14:30:40 CEST] <jookiyaya> [05:05] <jookiyaya> jkqxz why still use x264, when there is x265
[14:30:41 CEST] <jookiyaya> [05:06] <Spring> 1.7 too fast 1.5 too slow, 1.6 too slow... :/
[14:30:41 CEST] <jookiyaya> [05:06] <jookiyaya> spring too fast?
[14:30:41 CEST] <jookiyaya> [05:06] <jkqxz> x264 has better tuning to occupy the processor as much as possible, so it will be a better benchmark.  Also 9001 times as many people use it, which probably makes it more relevant...
[14:31:19 CEST] <jookiyaya> 05:07] <Spring> jookiyaya, can't find the correct calculation for atempo when stacking values over 2.0 (ie: 200%)
[14:31:19 CEST] <jookiyaya> [05:07] <jookiyaya> spring is that a typo
[14:31:19 CEST] <jookiyaya> [05:08] <furq> Spring: what final value do you want
[14:31:19 CEST] <jookiyaya> [05:08] <jookiyaya> spring, huh?  please explain why you said  1.7 = too fast
[14:31:19 CEST] <jookiyaya> [05:09] <furq> if you want 3.0 then use 2.0 and 1.5
[14:31:19 CEST] <jookiyaya> [05:10] <jookiyaya> furq do you have any idea what spring is saying
[14:31:56 CEST] <durandal_1707> He is attempting to slow video, IIRC
[14:32:35 CEST] <jookiyaya> spring    wait, when you said  too fast/slow  , you are not talking about encoding speed?
[14:33:33 CEST] <durandal_1707> no
[14:33:45 CEST] <jookiyaya> oh
[14:34:13 CEST] <jookiyaya> what does it mean  "too slow" then
[14:36:06 CEST] <durandal_1707> that video is out of sync I guess
[14:36:39 CEST] <Spring> video speed, anyway you can PM me if you'd like to continue the questions. Found that I set the multiplier too low for the batch script though (as it doesn't support decimal numbers), which may explain why the setpts value is off.
[14:37:07 CEST] <Spring> well, not natively decimal, I convert them
[14:40:15 CEST] <Spring> every number after the decimal is stripped so I have to multiply by like 100,000 to preserve enough of the value
[14:57:47 CEST] <flai> furq: Apparently it was the AAC. The encoded aac from OBS isn'
[14:58:07 CEST] <flai> doesn't work, if I reencode with -c:a aac it works
[14:58:11 CEST] <flai> Thanks for your help!
[16:40:50 CEST] <yangxuan8282_> hi
[16:42:24 CEST] <yangxuan8282_> when I building ffmpeg in the chroot, it says: ERROR: mmal not found.
[16:45:09 CEST] <yangxuan8282_> anybody face same issues?
[16:59:34 CEST] <furq> yangxuan8282_: check config.log
[16:59:58 CEST] <furq> https://github.com/FFmpeg/FFmpeg/blob/master/configure#L5766
[17:03:07 CEST] <yangxuan8282_> furq thanks for reply, the config.log is too long, I don't know check which part
[17:03:13 CEST] <furq> the very end
[17:14:32 CEST] <viric> Hello ffmpeg
[17:14:48 CEST] <durandal_1707> Hello
[17:14:48 CEST] <viric> is there anything I can use to adjust audio sync&drift between tracks?
[17:15:10 CEST] <viric> other than adelay
[17:15:52 CEST] <viric> I mean - I want help finding out the adelay parameter and speed adjustment :)
[17:16:16 CEST] <viric> (atempo)
[17:18:06 CEST] <durandal_1707> depends what tracks have
[17:18:50 CEST] <viric> they are independant; one from a videocamera, the other from a dictophone
[17:19:00 CEST] <viric> (MTS file + WMA file)
[17:19:37 CEST] <durandal_1707> and how they differ?
[17:19:52 CEST] <viric> could be 1 minute
[17:20:06 CEST] <viric> the drift of the wma may be 0.1s per hour
[17:20:25 CEST] <viric> maybe 0.5. 0.1 wouldn't be very much noticeable
[17:20:57 CEST] <viric> But the point is that I don't know anything else than testing adelay values with ffplay
[17:21:03 CEST] <viric> and checking manually the sync
[17:22:06 CEST] <durandal_1707> adelay is for constant difference
[17:23:01 CEST] <viric> adelay+atempo, I use
[17:23:23 CEST] <viric> but I don't want to *calculate* the values by manual brute force
[17:24:26 CEST] <durandal_1707> There is some kind of sound which indicate relative time
[17:24:30 CEST] <durandal_1707> ?
[17:24:36 CEST] <viric> it is speech
[17:24:38 CEST] <viric> I did not clap hands
[17:24:56 CEST] <viric> both tracks quite have the same speech; simply one is clearer
[17:25:30 CEST] <viric> I tried using kdenlive's "audio reference" and alignment, but it crashes. :)
[17:25:42 CEST] <viric> (and it is only about delay, not drift)
[17:27:12 CEST] <durandal_1707> Hmm, silencedetect?
[17:27:50 CEST] <durandal_1707> it will show when silence ended
[17:28:08 CEST] <durandal_1707> but nothing for drift
[17:28:56 CEST] <viric> I just feel that I'm missing something relevant; how can people compose tracks using multiple microphones?
[17:29:06 CEST] <durandal_1707> unless voice have beats and someone write beat detector
[17:29:57 CEST] <viric> audacity has something... http://manual.audacityteam.org/man/audio_alignment_tips.html
[17:30:05 CEST] <durandal_1707> viric: compensationdelay filter if you know distance
[17:30:25 CEST] <viric> what is compensationdelay?
[17:30:48 CEST] <durandal_1707> calf plugin ported to lavfi
[17:34:50 CEST] <viric> hm interesting
[17:34:56 CEST] <viric> well, I'm not really mixing the audio though
[17:35:03 CEST] <viric> I want dictaphone track aligned with video
[17:36:37 CEST] <durandal_1707> yea, I guess load tracks in audacity and see how much to shift
[17:37:18 CEST] <viric> Another thing I have to do most often is to switch the microphone during a talk of two people
[17:37:24 CEST] <viric> imagine each has its own good microphone
[17:37:47 CEST] <viric> I'm trying with kdenlive.. but it is a hell to do that. Maybe I just don't know a term for it to google for...
[17:38:11 CEST] <bencoh> audacity is probably handy for that yeah
[17:38:16 CEST] <bencoh> you'll just "see" it
[17:38:21 CEST] <bencoh> (the waves)
[17:38:37 CEST] <viric> aha
[17:38:52 CEST] <viric> I should try.
[17:39:09 CEST] <viric> So if I extract the audio from the mts and load it in audacity... the startpts would be correct once I mix them with video again?
[17:40:46 CEST] <durandal_1707> no you would need to insert samples or remove them...
[17:41:27 CEST] <viric> I'm a bit disturbed by this mts specificalities. Looking at ffprobe -show_streams I see for stream 0 (video):
[17:41:30 CEST] <viric> start_pts=95400
[17:41:30 CEST] <viric> start_time=1.060000
[17:41:37 CEST] <viric> And for the audio track of the mts:
[17:41:42 CEST] <viric> start_pts=91800
[17:41:42 CEST] <viric> start_time=1.020000
[17:42:38 CEST] <durandal_1707> that is another issue
[17:42:51 CEST] <viric> :)
[17:44:07 CEST] <viric> so if I extract the audio into a wav, I'd lose the start_time/pts, isn't it?
[17:44:36 CEST] <viric> and direct remix would have a 1s a/v misalignment
[17:45:59 CEST] <viric> Moreover, I have *two* mts that are meant to be concatenated. One has start_time=1.06, the other start_time=7019.74
[17:46:30 CEST] <viric> how can I use ffmpeg so the two files look like a single recording in input?
[17:47:56 CEST] <durandal_1707> there's concat demuxer and filter
[17:48:33 CEST] <viric> but will that pay attention to those start_time/start_pts?
[17:51:10 CEST] <durandal_1707> I guess, at least demuxer
[17:53:17 CEST] <viric> kdenlive makes a short quirk on file switch, while concatenating
[18:35:52 CEST] <yangxuan8282_> @furq I read the config.log, the last part is: fatal error: interface/mmal/mmal.h: No such file or directory
[18:37:43 CEST] <yangxuan8282_> then I copy the mmal directory to the chroot: sudo cp -a /opt/vc/* /home/alarm/chroot/alarm/opt/ , but still got same issues
[19:42:10 CEST] <ozette> what's the mimetype of hls segmented videos .m3u(8)?
[19:46:10 CEST] <c_14> Apparently application/vnd.apple.mpegurl
[19:46:14 CEST] <c_14> https://www.ietf.org/id/draft-pantos-http-live-streaming-19.txt section 10
[19:47:13 CEST] <ozette> hmm
[19:47:27 CEST] <ozette> what about application/x-mpegURL ?
[19:47:35 CEST] <ozette> i assume both will work
[19:49:17 CEST] <c_14> According to the draft that's not valid.
[19:50:38 CEST] <c_14> It's either application/vnd.apple.mpegurl or audio/mpegurl
[19:52:56 CEST] <c_14> There may, however, be legacy reasons why that would be supported by some applications.
[19:55:24 CEST] <ozette> hm ok
[20:17:14 CEST] <jookiyaya> big news: zen destroys intel  http://static2.gamespot.com/uploads/scale_super/460/4601080/3046688-8311357204-AMD-Z.jpg
[20:20:42 CEST] <Chloe> There are three types of lies; lies, damned lies, and benchmarks.
[20:21:16 CEST] <jookiyaya> so benchmark scores mean nothing?
[20:22:23 CEST] <c_14> Never trust any statistic you didn't manipulate yourself.
[20:23:37 CEST] <Chloe> jookiyaya: they mean nothing, and you probably shouldnt post them for a third time.
[20:24:08 CEST] <jookiyaya> chloe of course they mean something
[20:31:32 CEST] <iive> jookiyaya: you already discussed that news here
[20:32:13 CEST] <jookiyaya> iive i don't even remember you ever commenting on it
[20:39:18 CEST] <DHE> amd's been behind intel for a long time now, how'd they catch up so quick?
[20:42:45 CEST] <jookiyaya> i wouldn't say quick though
[20:43:12 CEST] <iive> DHE: look the discussion from 6-7 hours ago.
[20:53:18 CEST] <DHE> iive: will do later, gotta find if I have an ffmpeg bug or if I've screwed something up...
[21:34:26 CEST] <DHE> av_interleave_write_frame() is supposed to buffer packets from all streams such that 'ffprobe -show_frames' would show all frames in strictly increasing DTS order to the whole file, not just each independent stream. is that right?
[21:40:39 CEST] <DHE> the reason I ask is I need output that is so "cleaned" but ffmpeg isn't give it to me
[22:00:30 CEST] <DHE> am I wrong?
[22:37:46 CEST] <chris_> Hi! Can someone please help me with h264_nvenc?
[22:38:04 CEST] <DHE> that's not how it works. state your question
[22:39:12 CEST] <chris_> I get this error when using the encoder: [h264_nvenc @ 0x260a180] Failed to create nvenc instance: invalid version (15)
[22:40:27 CEST] <DHE> you might be using an old (or new?) version of the SDK
[22:40:51 CEST] <DHE> or an old/new version of ffmpeg. either way there was a version bump in the last few months
[22:41:40 CEST] <chris_> It's supposed to be v 7.0.
[22:42:07 CEST] <chris_> Just compiled today.
[22:42:40 CEST] <chris_> with latest sources.
[22:43:04 CEST] <viric> chris_: the nvenc headers you used may not match your GL driver version
[22:43:45 CEST] <chris_> Ok. But it still compiles, though?
[22:44:05 CEST] <DHE> it'll compile, but not run
[22:44:32 CEST] <viric> chris_: the headers only describe the lib that will be used at runtime (your nvidia driver)
[22:44:44 CEST] <viric> you are responsible to match the headers with the nvidia driver you run
[22:45:34 CEST] <chris_> That makes sense, but sure how to accomplis that. I'm just a noob on coding.
[22:45:41 CEST] <chris_> not sure
[22:45:51 CEST] <viric> if you used the latest headers, just update your driver
[22:51:10 CEST] <chris_> My nvidia driver is v 361.42. Is ther an "apt-get install nvidia-latest" that get the latest driver?
[22:55:34 CEST] <viric> that sounds debianish. No idea.
[22:57:09 CEST] <DHE> no, that's how nvidia numbers their drivers.
[22:57:19 CEST] <DHE> though it is a little out of date now
[23:04:28 CEST] <Prelude2004c> hey guys.. good day ... i am looking for some help here with something silly that i can't figure out... so i am splitting an input like " -filter_complex '[0:v]yadif,format=nv12,split=3[vout1][vout2][vout3]" and the output can't keep up with the input and its running at less than 1x... ok so to debug i did -filter_complex '[0:v]yadif,format=nv12,split=1[out1]   and it did the same thing..only a single output . same problem. now when i
[23:04:28 CEST] <Prelude2004c> took out the ,split=1[out1] -map '[out1]', everything went ok and it was 1x .. any ideas ?
[23:17:26 CEST] <theTroy> Hi! I am trying to encode with hevc_nvenc codec - it works fine and does use the nvidia card. Unfortunately the resulting video is identified as 8-bit uncompressed RGB by ubuntu and is not played back by totem/vlc at all?
[23:20:47 CEST] <Prelude2004c> theTroy, try nvenc_hevc .. it may be defaulting to a different encoder because you have it backwards
[23:21:42 CEST] <theTroy> Prelude2004c: sorry I did the right way round - written it wrong here. My command line is : ffmpeg -i INPUT -c:v nvenc_hevc  -profile high444p -pixel_format yuv444p -preset default -b:v 3000k -c:a copy   OUTPUT
[23:21:42 CEST] <theTroy> ]]]
[23:21:50 CEST] <theTroy> apologies for the brackets
[23:22:02 CEST] <Prelude2004c> that is the output format you set ?
[23:22:07 CEST] <Prelude2004c> the -f
[23:22:20 CEST] <theTroy> I dont have -f doh! thanks!
[23:23:10 CEST] <theTroy> Prelude2004c: although here they do not suggest to use -f https://trac.ffmpeg.org/wiki/Encode/H.264
[23:23:26 CEST] <Prelude2004c> maybe... it was a suggestion :)
[23:23:39 CEST] <theTroy> right, encoding now - we will see, the output was set as .avi
[23:23:58 CEST] <Prelude2004c> maybe its not detecting the .avi
[23:24:04 CEST] <Prelude2004c> so its picking some other default
[23:24:12 CEST] <DHE> you don't want avi. ever
[23:26:00 CEST] <Prelude2004c> DHE, you know this stuff well....
[23:26:04 CEST] <Prelude2004c> hey guys.. good day ... i am looking for some help here with something silly that i can't figure out... so i am splitting an input like " -filter_complex '[0:v]yadif,format=nv12,split=3[vout1][vout2][vout3]" and the output can't keep up with the input and its running at less than 1x... ok so to debug i did -filter_complex '[0:v]yadif,format=nv12,split=1[out1]   and it did the same thing..only a single output . same problem. now when i
[23:26:04 CEST] <Prelude2004c> took out the ,split=1[out1] -map '[out1]', everything went ok and it was 1x .. any ideas ?
[23:26:09 CEST] <Prelude2004c> i am stuck on that one :(
[23:26:37 CEST] <Prelude2004c> wonder if some bug i the code.. why would split=x[] actually slow down the output ( or input )
[23:26:40 CEST] <Prelude2004c> even when =1
[23:27:29 CEST] <DHE> you got multithreading enabled?
[23:27:42 CEST] <Prelude2004c> yup.. -threads 0
[23:27:49 CEST] <Prelude2004c> you mean on the ffmpeg itself?
[23:28:08 CEST] <Prelude2004c> --enable-pthreads yes i do
[23:28:53 CEST] <Prelude2004c> strange thing is.. even 1 input causes an issue... when i don't use the split , works perfectly
[23:29:07 CEST] <Prelude2004c> problem is i have to output 3 different bit rates
[23:32:19 CEST] <theTroy> So yeah nvenc_hevc shows up as 24 bits RGB (RV24) in vlc when played back
[23:32:25 CEST] <theTroy> and it actually plays back no video whatsoever
[23:32:46 CEST] <theTroy> if I encode with non nvenc h265 with libav - it does play it back
[23:34:22 CEST] <Prelude2004c> sorry " theTroy" i am still learning too
[23:34:59 CEST] <theTroy> yeah np Prelude2004c hopefully someone else would be able to help
[00:00:00 CEST] --- Mon Aug  8 2016



More information about the Ffmpeg-devel-irc mailing list