[Ffmpeg-devel-irc] ffmpeg.log.20170418

[burek]

[00:00:09 CEST] <erick3k> gonna test it you saved me thank you so much
[00:00:26 CEST] <furq> if you need multiple files then you can remux them all to ts and cat src1.ts src2.ts src3.ts
[00:00:29 CEST] <furq> etc
[00:00:34 CEST] <furq> or just cat *.ts if they're in alphabetical order
[00:00:43 CEST] <erick3k> nice
[00:00:59 CEST] <furq> it'll only work with ts and maybe other streamable containers
[00:01:16 CEST] <furq> mp4 is not one of those
[00:01:20 CEST] <erick3k> got you
[00:02:29 CEST] <kerio> is there a rawvideo format that supports timing?
[00:02:47 CEST] <furq> y4m?
[00:03:19 CEST] <kerio> that's framerate tho
[00:03:21 CEST] <aconz2> I'm getting an audio sync problem using the loudnorm filter. demonstration here: https://pastebin.com/RFE7Hd9u -- the output loses the proper start_time
[00:03:22 CEST] <kerio> i want variable framerate
[00:03:25 CEST] <furq> oh
[00:03:28 CEST] <furq> no idea then
[00:03:40 CEST] <erick3k> furq tried ffmpeg -i small2.mp4 -c copy source.ts but got error https://0bin.net/paste/5CexxYFe1Pns5Sap#ayOhkyL2rRTgS7KXoE4Nu+MsmWu7Ceg6HfjwHj2ujPG
[00:04:00 CEST] <furq> -bsf:v h264_mp4toannexb
[00:04:18 CEST] <kerio> furq: ideally with microsecond granularity, for perfect compatibility with ttyrecs
[00:04:43 CEST] <furq> erick3k: also you might as well stream the audio without reencoding, it's already aac-lc
[00:05:08 CEST] <erick3k> ok
[00:05:10 CEST] <furq> replace the entire line containing -acodec with -c:a copy
[00:07:03 CEST] <erick3k> that will save some cpu?
[00:07:12 CEST] <furq> yeah and the audio will sound better
[00:12:45 CEST] <erick3k> furq sorry to bother, tried that -c:a and get an flv error https://0bin.net/paste/wuLUeg0cYUujf+zW#lOK6DEwB5m8iZgvo6hW6jIhyqjwoK1OYqqmMN5ssbha
[00:13:04 CEST] <furq> pastebin the command
[00:14:04 CEST] <erick3k> https://0bin.net/paste/wuLUeg0cYUujf+zW#lOK6DEwB5m8iZgvo6hW6jIhyqjwoK1OYqqmMN5ssbha
[00:14:21 CEST] <furq> the command, not the output
[00:16:21 CEST] <erick3k> https://0bin.net/paste/1U+jbBwwrrzSduJf#n2564NftzOIlPj+3IfvDrUgDO4w1PFnXtOUw2ZZduhT
[00:16:31 CEST] <furq> -c:a copy
[00:17:22 CEST] <erick3k> add "copy" to it?
[00:19:18 CEST] <erick3k> says to add -bsf:a aac_adtstoasc
[00:19:20 CEST] <erick3k> should i?
[00:19:22 CEST] <furq> yes
[00:19:59 CEST] <erick3k> cool
[00:23:22 CEST] <BtbN> shouldn't that happen automatically by now?
[00:24:49 CEST] <furq> not in 2.8
[00:26:09 CEST] <erick3k> furq would it stream at 720p or would it go down to 480p?
[00:27:56 CEST] <erick3k> does converting it to ts reduces the video size?
[00:32:40 CEST] <furq> no
[00:33:36 CEST] <erick3k> hum doesn't show on youtube anymore
[00:33:46 CEST] <erick3k> 720p was showing before :(
[00:35:47 CEST] <erick3k> maybe because of buffsize?
[00:37:52 CEST] <erick3k> furq any idea?
[00:46:04 CEST] <erick3k> nvm seems like it was the source
[00:46:09 CEST] <erick3k> sound is horrible without converting do
[00:46:13 CEST] <erick3k> like a tick on the sound
[00:59:25 CEST] <Guest36915> Hello guys please me help, avoid this errors:
[00:59:28 CEST] <Guest36915> https://pastebin.com/Mnrj923B
[00:59:57 CEST] <BtbN> fix the stream
[01:00:27 CEST] <Guest36915> BtbN, so its stream problem ?
[01:01:12 CEST] <Guest36915> No way fix from ffmpeg ?
[01:01:33 CEST] <BtbN> that looks like the stream is utter garbage, yes
[01:02:03 CEST] <Guest36915> thanks!
[01:14:15 CEST] <iive> Guest36915: udp transport is very sensitive to delivery order
[01:24:11 CEST] <dystopia_> Past duration 0.999443 too large= 3455541kB time=01:17:38.52 bitrate=6076.6kbits/s dup=0 drop=6 speed=0.565x
[01:24:11 CEST] <dystopia_>     Last message repeated 11 times
[01:24:18 CEST] <dystopia_> what do these warnings mean?
[01:30:16 CEST] <iive> dystopia_: you write to .mp4 ?
[01:32:49 CEST] <dystopia_> it's going to mkv
[01:39:24 CEST] <Threads> dystopia_ are you cutting(copy) to mkv or encoding when getting this problem
[01:41:11 CEST] <Threads> i would say you could be dropping frames when encoding
[01:50:21 CEST] <Guest36915> iive, thank you
[01:50:44 CEST] <dystopia_> it's encoding to x264 threads
[01:51:10 CEST] <dystopia_> it's an mpeg2 program stream source
[01:51:22 CEST] <dystopia_> i only every see those errors on 1 channel i record from
[01:51:39 CEST] <furq> i take it this isn't a live source
[01:52:31 CEST] <Guest36915> Gus, during transcoding stream i have something like this: frame=1000 fps=25 q=34.....    So i want to know what's means q?  I didn't found in documentation
[01:52:47 CEST] <furq> q is quantiser
[01:52:59 CEST] <furq> it's not something you need to pay attention to
[01:53:30 CEST] <Guest36915> furq, is it related whth CPU ?
[01:53:37 CEST] <furq> no
[01:53:47 CEST] <Guest36915> furq, thank you
[01:56:21 CEST] <dystopia_> it's not live furq no, it's pre recorded with dvbviewer
[02:37:01 CEST] <AndrewMock> What's an easy to test A/V sync?
[02:37:25 CEST] <AndrewMock> any sort of input device that's builtin for that?
[02:37:36 CEST] <AndrewMock> I think my video is encoding too slowly
[03:23:00 CEST] <kepstin> dystopia_: that error means that ffmpeg's estimation of the input framerate wasn't quite correct, and the stream was actually vfr. I dunno any really good way to fix it, but if you know the framerate the video is supposed to be, sticking a 'fps' filter in helps. Using "-vsync vfr" might work too?
[03:24:42 CEST] <dystopia_> it's cfr 25fps interlaced that i bob deinterlace to 50fps
[03:24:54 CEST] <dystopia_> i will run a test with fps and see what happens
[03:27:29 CEST] <gurki> there is such a thing as vfr? :S
[03:27:51 CEST] <gurki> how is that better than jumping from one keyframe to another?
[03:30:51 CEST] <kepstin> gurki: actual use of vfr is kinda rare, but 'matroska' format doesn't actually really support cfr - frame timestamps are just rounded to the nearest 1/1000s :/
[03:32:26 CEST] <gurki> yay :/
[03:32:44 CEST] <kepstin> the main semi-legitimate use of vfr that i've seen is handling ntsc content that's a mix of 60i (interlaced) and 60i (telecined from 24p), where it can be sometimes turned into sections of 30p or 60p by deinterlacing plus other sections of 24p by detelecining
[05:15:34 CEST] <Fenrirthviti>  warning: Truncating likely oversized PPS (zu > zu) <- the heck does that mean
[05:17:56 CEST] <Fenrirthviti> https://ffmpeg.org/pipermail/ffmpeg-devel/2016-June/195394.html is all I can find on that, and it doesn't look like they ever opened a ticket
[05:19:10 CEST] <Fenrirthviti> Ah wait maybe they did, just no activity on it? or another user opened https://trac.ffmpeg.org/ticket/5731
[07:45:34 CEST] <mappy> how does ffmpeg 3.3 internal opus encoder compare to libopus encoder?
[08:50:38 CEST] <damdai> why does so many players have problem playingback  WMV
[08:54:27 CEST] <durandal_1707> damdai: whats an issue?
[08:54:47 CEST] <damdai> playing back wmv
[08:56:03 CEST] <durandal_1707> damdai: what exactly?
[08:56:11 CEST] <damdai> for example, it takes forever to load
[08:56:17 CEST] <damdai> or forever to seek
[08:56:50 CEST] <durandal_1707> with files?
[08:56:52 CEST] <damdai> i wonder if i mux to mkv, it fixes everthing
[08:56:58 CEST] <damdai> durandal_1707 yes
[08:57:57 CEST] <durandal_1707> damdai: you could open bug report and provide link to samples
[08:58:34 CEST] <damdai> does mpc-hc use ffmpeg to playback wmv
[09:00:27 CEST] <durandal_1707> i doubt they use libavformat
[09:00:43 CEST] <durandal_1707> did you tried ffplay?
[09:01:23 CEST] <damdai> no
[09:01:30 CEST] <damdai> let me try ffplay
[09:05:07 CEST] <damdai> durandal_1707 ffplay is good
[09:06:14 CEST] <durandal_1707> than its mpc-hc issue
[09:06:47 CEST] <damdai> does vlc not use ffmpeg ?
[09:07:44 CEST] <durandal_1707> depends
[09:07:54 CEST] <durandal_1707> it can use libav
[09:08:06 CEST] <durandal_1707> and have own demuxers
[09:08:19 CEST] <damdai> any idea what mpc-hc use
[09:08:51 CEST] <durandal_1707> internal asf implementation
[09:09:36 CEST] <damdai> smplayer plays wmv ok too
[09:09:57 CEST] <damdai> but vlc/mpchc plays them horribly
[09:10:38 CEST] <damdai> WMP plays wmv the best, but WMP makes harddrive noise whenever i open WMV,
[09:10:44 CEST] <damdai> WMP plays wmv the best, but WMP makes harddrive noise whenever i open WMP
[09:10:53 CEST] <durandal_1707> ask vlc/mpchc devs
[09:43:14 CEST] <axeltory> /list
[10:39:13 CEST] <axeltory> Hello everyone. Could anyone help me to understand a thing about FFMPEG. Do hardware acceleration methods "dxva2" and/or "cuvid" mean that FFMPEG will use NVIDIA GPUs?
[10:40:21 CEST] <axeltory> I compile FFMPEG from sorce in Windows 10 using MinGW msys command prompt
[10:40:46 CEST] <tdr> https://trac.ffmpeg.org/wiki/HWAccelIntro
[10:41:47 CEST] <axeltory> CUDA is also known as CUVID. Thank you!
[13:29:42 CEST] <superware> the mp4 output format returns "Could not find tag for codec none in stream #1, codec not currently supported in container" for avformat_write_header, stream #1 is a AVMEDIA_TYPE_DATA with codec AV_CODEC_ID_NONE, any ideas?
[13:54:57 CEST] <dystopia_> the codec isnt yet supported by ffmpeg superwar
[13:54:59 CEST] <dystopia_> https://ffmpeg.org/bugreports.html
[13:55:09 CEST] <dystopia_> upload a sample of it and maybe they will add support
[13:57:05 CEST] <superwar> dystopia_: but AV_CODEC_ID_NONE isn't a codec...
[13:57:16 CEST] <superwar> it's a "none" codec :)
[13:57:59 CEST] <durandal_1707> its for unknown codecs
[14:01:36 CEST] <superware> durandal_1707: no, I think AVMEDIA_TYPE_UNKNOWN is for unknown codecs, not AV_CODEC_ID_NONE
[14:01:43 CEST] <superware> oops, sorry
[14:02:46 CEST] <durandal_1707> first one is for unknown TYPE
[14:03:43 CEST] <durandal_1707> make sure its not font
[16:15:40 CEST] <Krstjns> So heres something... I am doing video decoding and encoding with ffmpeg.c, while trying to make sure that the video timings remain the same in the final result.. however I am noticing that ffprobe gives me different fps (29.97 vs 30.06) while tbr and tbc remain the same
[16:15:55 CEST] <Krstjns> What could i be missing in the encoding process?
[16:21:45 CEST] <Krstjns> this is not done via command line tool but the C library
[16:23:04 CEST] <kepstin> alright, as a start, what container are you using for input and output?
[16:23:35 CEST] <Krstjns> This is the setup for encoders video stream: https://pastebin.com/JrTS2Crr
[16:24:58 CEST] <kepstin> yes, but what container are you putting it into? :)
[16:25:22 CEST] <Krstjns> Umm, what do you mean by container
[16:25:39 CEST] <kepstin> are you saving the video to a file on disk? what AVFormat? streaming?
[16:25:59 CEST] <Krstjns> ah, both input and output are on dick
[16:26:03 CEST] <Krstjns> disk*...
[16:26:06 CEST] <kepstin> also, you might want to set stream->avg_frame_rate to the expected frame rate of the stream, that can give a hint to the muxer in some cases.
[16:26:06 CEST] <Krstjns> ...
[16:26:22 CEST] <kepstin> ok, what file format? matroska? mpeg-ts? something else?
[16:26:33 CEST] <Krstjns> all is mp4/h264
[16:26:44 CEST] <kepstin> mp4, ok. that's what I wanted to know :)
[16:28:37 CEST] <kepstin> does the file play back at the correct rate (i.e. is this only a cosmetic problem that the wrong framerate is being probed?)
[16:29:06 CEST] <Krstjns> the playback is correct
[16:29:57 CEST] <Krstjns> Or so it seems, 30.06 and 29.97 migh be hard to tell apart
[16:30:31 CEST] <Krstjns> Im seeing this as a symptom, I am getting problems with video frames later down the pipeline
[16:30:35 CEST] <kepstin> Krstjns: ok, then I suggest just setting stream->avg_frame_rate when encoding, that should fix the metadata in the container
[16:30:58 CEST] <Krstjns> Ill give that a shot
[16:34:06 CEST] <Krstjns> Sadly that doesnt seem to help
[16:34:16 CEST] <Krstjns> Let me define the issue a bit more specificaly
[16:37:33 CEST] <Krstjns> when reading the source video the timings I get are: https://pastebin.com/tAvR5xWm
[16:39:18 CEST] <Krstjns> upon encoding the same readings give me: https://pastebin.com/HhfPkzyL
[16:39:42 CEST] <Krstjns> the time_base seems completely off
[16:40:00 CEST] <Krstjns> even tho I do specify it to be 30000 / 1001
[16:41:24 CEST] <kepstin> so, FPS_NUMERATOR and FPS_DENOMINATOR in your code are 30000 and 1001?
[16:41:29 CEST] <Krstjns> yes
[16:42:39 CEST] <Krstjns> I am also using h264_nvenc for encoding if that gives any hints
[16:42:50 CEST] <Krstjns> wouldnt be the first problem we have with it :/
[16:43:17 CEST] <kepstin> well, you could check what the pts looks like both before and after passing it to the encoder, in the avframe, I guess?
[16:48:02 CEST] <alex88> is there a way to tell ffmpeg to use a fixed fps instead of using probing?
[16:48:34 CEST] <Krstjns> the pts on AVFrames are simply set to 1.. 2... 3.. and so on.. huh
[16:49:11 CEST] <Krstjns> yet after decoding it still gives me values of 1001*x
[16:51:29 CEST] <kepstin> alex88: with the ffmpeg cli? use -r <fps) before the input
[16:51:36 CEST] <kepstin> -r <fps>
[16:52:16 CEST] <alex88> kepstin: ok thank you!
[16:52:29 CEST] <alex88> I'm trying to get a stream from an rtsp server and I randomly get issue
[16:53:08 CEST] <kepstin> alex88: that option will cause it to ignore the fps from the source, and write new timestamps corresponding to a constant fps stream.
[16:53:21 CEST] <alex88> sometimes is like the first file, sometimes like the second https://gist.github.com/alex88/e20385224d98d367ce868e8a29d943a1
[16:53:37 CEST] <alex88> and sometimes after a while it starts writing to file
[16:53:56 CEST] <kepstin> alex88: well, that's normal, it can only start at a keyframe, so it has to wait for a keyframe to show up. nothing to do with timestamps
[16:54:07 CEST] <alex88> oh ok, thank you
[16:54:21 CEST] <alex88> however with -r it started right after launching the command
[16:54:22 CEST] <kepstin> (if you have control of the source, you should reduce the key frame interval/gop size)
[16:56:01 CEST] <alex88> kepstin: I'll try, I'm using webcam -> v4l2loopback -> v4l2compress_h264 -> v4l2rtspserver -> ffmpeg
[16:56:45 CEST] <Krstjns> huh, se heres a fun thing to add: going with software encoding instead of hardware gives me 59.94 tbc instead of 29.97
[16:56:55 CEST] <Krstjns> with still messed up fps
[23:03:48 CEST] <m8377> I've got a batch script setup to scale mp4 videos in a folder to 1920x1080. Can ffmpeg pick up on videos that are already the proper resolution and skip them?
[23:05:48 CEST] <Tatsh> since you're in windows, consider using powershell instead; and then the grep aliased cmdlet
[23:06:02 CEST] <Tatsh> ffmpeg -i <filename> | grep 1920
[23:07:53 CEST] <Tatsh> or in batch, use findstr
[23:10:25 CEST] <m8377> Tatsh good idea
[23:13:13 CEST] <furq> m8377: http://vpaste.net/X44pI
[23:20:23 CEST] <utack> is there a windows build for the new ffmpeg version yet?
[23:25:42 CEST] <BtbN> There is a windows build for almost every commit, or at least one every day.
[23:28:42 CEST] <utack> BtbN oh where. i only know about the one linked on the download page
[23:28:52 CEST] <BtbN> yes, exactly there.
[23:29:25 CEST] <Tatsh> i'm trying to do a 5.0 (5.1 without subwoofer) file from stereo
[23:29:27 CEST] <utack> ah ok. i will wait for that then. thanks
[23:29:35 CEST] <Tatsh> i specified -i 5 times with FL, FR, etc
[23:29:37 CEST] <Tatsh> i got this line
[23:29:44 CEST] <Tatsh>     Stream #0:1: Audio: aac (LC) ([255][0][0][0] / 0x00FF), 48000 Hz, 5.0, fltp, 325 kb/s (default)
[23:29:50 CEST] <BtbN> There's nothing to wait for, there's at least one build per day, if not even more frequently.
[23:30:01 CEST] <Tatsh> is 5.0 '5.1 without LFE' ?
[23:30:37 CEST] <Tatsh> i haven't seen an example where i can definitely specify where each channel goes
[23:30:47 CEST] <BtbN> hm, looks like it slowed down, or broke. At least the Zeranoe ones. Maybe the same dacklink issue
[23:31:13 CEST] <utack> http://ffmpeg.zeranoe.com/builds/win64/static/
[23:31:15 CEST] <Tatsh> under ffmpeg -layouts, it says 5.0            FL+FR+FC+BL+BR
[23:31:17 CEST] <utack> not since the 11th
[23:31:21 CEST] <Tatsh> so it seems correct :)
[23:31:43 CEST] <Tatsh> my command is ffmpeg -i gits.mkv -i left.wav -i right.wav -i cfe.wav -i back-left.wav -i back-right.wav -filter_complex '[1:a][2:a][3:a][4:a][5:a]amerge=inputs=5[aout]' -map 0:0 -map '[aout]' -c:a aac -c:v copy out.mkv
[23:31:57 CEST] <Tatsh> will have to test in my system
[23:38:50 CEST] <Tatsh> well, i played the file and it works as expected
[23:39:21 CEST] <BtbN> utack, that's still more recent than 3.3 though
[23:40:10 CEST] <utack> ah, i see. unfortunately libvpx 1.6.1, i was hoping to try 1.6.2 with the new multithreading mode. guess i will just have to wait for a bit
[23:40:34 CEST] <BtbN> well, the libvpx version used is entirely up to the one making the build
[23:40:35 CEST] <utack> oh it is not officially released yet...
[23:40:55 CEST] <furq> did they finally add frame threading
[23:41:01 CEST] <BtbN> 3.3 is tagged, I'd consider that a release
[23:41:08 CEST] <furq> he means libvpx 1.6.2
[23:41:21 CEST] <utack> furq https://www.phoronix.com/scan.php?page=news_item&px=VP9-Faster-MT
[23:41:27 CEST] <furq> oh, row threading
[23:41:35 CEST] <utack> so no
[23:41:42 CEST] <furq> i guess it's better than nothing
[23:42:16 CEST] <utack> i really hope they fix that for av1...the state of the reference encoder compared to x264 and x265 is really poor unless you are some advanced user who has a full framework to split input into blocks to separately encode at once
[23:42:29 CEST] <furq> i'm told it's planned for av1
[23:43:04 CEST] <utack> that would be nice, yes
[23:44:02 CEST] <utack> https://bugs.chromium.org/p/aomedia/issues/detail?id=18&q=parallel
[23:45:57 CEST] <utack> it is probably lowish on their priority list, because they need to fix the format so hardware partners can start working on it?
[00:00:00 CEST] --- Wed Apr 19 2017