[Ffmpeg-devel-irc] ffmpeg.log.20140208

Sun Feb 9 02:05:01 CET 2014

[00:45] <luc4> Hello! I'm experiencing a little problem transcoding a video file. When trying to transcode I get "Frame rate very high for a muxer not efficiently supporting it." and "MB rate (324000000) > level limit (2073600)", and in fact the transcoding seems to never end. But ffprobe returns this for my stream "Stream #0:0(eng): Video: h264 (Baseline) (avc1 / 0x31637661), yuv420p, 1280x720, 7569 kb/s, SAR 65536:65536 DAR 16:9, 22.70 fps,
[00:45] <luc4> tbr, 90k tbn, 180k tbc (default)". Isn't this weird?
[01:04] <luc4> Hello71: http://pastebin.kde.org/pjw0kigkc
[01:07] <llogan> luc4: can ypu provide the input file?
[01:07] <llogan> you can PM the link if it's something you don't want in public
[01:12] <luc4> llogan: maybe I can cut it, let me try.
[01:13] <llogan> isn't it just ~62MB?
[01:18] <luc4> llogan: yes, but I can't provide it because the content cannot be disclosed unfortunately. I cut the first 5 seconds and I still see the same behavior. Can it be useful to reproduce the issue?
[01:18] <luc4> llogan: assuming this is an issue...
[01:22] <llogan> luc4: i'm guessing it may be useful. i can try to take a look at it later today.
[01:23] <luc4> llogan: do you want a link to download or a bug report?
[01:23] <llogan> just a link for now. i may not get to monkey with it much though until later
[01:24] <luc4> llogan: no problem, just a second.
[01:35] <llogan> if any of you here are students we are going to apply to Google Summer of Code this year if you're interested.
[01:36] <klaxa> what exactly does that entail? is it a lot of time per week?
[01:37] <llogan> it depends on the project you choose, your skill level, and your familiarity with FFmpeg, C, git, etc.
[01:37] <llogan> also, students get money
[01:38] <klaxa> that sounds really appealing
[01:38] <llogan> http://www.google-melange.com/gsoc/homepage/google/gsoc2014
[01:38] <llogan> or course we, as a project, have to get accepted first
[01:38] <llogan> s/or/of
[01:38] <klaxa> the fear i have is that i will not be able to provide as much worktime as i would like
[01:39] <llogan> http://wiki.multimedia.cx/index.php?title=FFmpeg_Summer_of_Code_2014
[01:39] <klaxa> apparently by changing the study system to the bachelor-master thing, students have less spare time. i may have no comparison to how it was prior, but my personal projects get less and less of my time
[01:40] <llogan> what's the "bachelor-master thing"?
[01:40] <klaxa> i'm lacking the technical terms here... like you get a bachelor's degree first and then a master's degree
[01:41] <klaxa> schedules are way tighter than they used to be
[01:41] <klaxa> ah, i'm in germany
[01:41] <klaxa> that is important to understand that i think
[01:41] <llogan> oh. also i haven't attended higher education for 10 years or so.
[01:42] <llogan> ...that decade was quicker than i expected.
[01:42] <klaxa> heh
[02:17] <shadowing> does anyone here have experience in re-encoding a ts file into another ts file?
[02:17] <shadowing> I'm trying to re-encode a stream of TS segments individually, then concatenate them together
[02:17] <shadowing> so far, I have the individual segment encoding done correctly
[02:18] <shadowing> but when I concatenate them to see if they'll play correctly together
[02:18] <shadowing> the video plays fine, but the audio has a noticeable stuttering at the segment boundaries
[02:22] <shadowing> I'd guess that when I encode each TS segment, some excess audio is being cut off
[02:22] <shadowing> anyone know of a way I can circumvent this?
[02:23] <shadowing> here's a pastebin of what I ran: http://pastebin.com/QUcZuRjB
[02:49] <kutemo> i have an issue over a filter that does not work if i resize
[02:50] <kutemo> if i use this -> ffmpeg -i input.m2ts -map 0:0 -vf yadif=1 -vcodec libx264 output.mp4
[02:50] <kutemo> yadif works, but if i resize it does not
[02:51] <kutemo> like so -> ffmpeg -i input.m2ts -map 0:0 -vf yadif=1 -vcodec libx264 -s 1280:720 output.mp4
[02:51] <kutemo> am i doing it wrong? why would it work if i don't resize though?
[02:52] <kutemo> btw, this is on ubuntu x64, ffmpeg version 0.10.9
[03:01] <kutemo> the output messages are irrelevant, because they don't show any issues but the quality is what's different
[03:01] <znf> kutemo, first of all, you're using "avconv" and not ffmpeg
[03:02] <kutemo> i'll upload two images with the details
[03:02] <znf> so try it with ffmpeg and not avconv
[03:02] <llogan> kutemo: the output messages are usually relevant
[03:02] <znf> as ffmpeg is at version 2.1.3, 0.10.9 sounds like avconv version ^^
[03:02] <llogan> although users have been resistant to providing the lately resulting in wasted time
[03:02] <llogan> *them (as in the console outputs)
[03:03] <llogan> and of course the console output indicates if you are actually using something from FFmpeg or not.
[03:03] <llogan> and since this is #ffmpeg we can only support FFmpeg stuff here
[03:03] <kutemo> znf, no i'm using ffmpeg. avconv is linked up to ffmpeg
[03:04] <llogan> see the link
[03:04] <kutemo> llogan, i did this a while ago and don't have the console ouputs anymore
[03:04] <znf> kutemo, don't confuse avconv with libav
[03:04] <znf> 0.10.9 is avconv version
[03:05] <llogan> there was an old 0.10.9 ffmpeg. perhaps he is using the PPA from Jon S.
[03:05] <kutemo> i remember the output messages and they were saying that both were using yadif
[03:06] <llogan> if you're filtering then use the scale filter instead of -s. that way you can control when exactly you want to scale by placing it whever you want in the filtergraph
[03:06] <llogan> -vf "yadif=1,scale=1280:-1"
[03:07] <kutemo> this is what i'm using -> https://launchpad.net/~jon-severinsson/+archive/ffmpeg/+packages
[03:07] <llogan> that's ffmpeg from FFmpeg, but it is old.
[03:10] <kutemo> ok, thanks llogan i'll give that a try
[03:25] <vl4kn0> Hi, I was looking over the decoding examples in ffmpeg git repository and the decoding_encoding.c examples includes a note in video decoding example saying: http://fpaste.org/75431/91826278/
[03:26] <vl4kn0> Now, the example only shows example of stream based codecs, but where can I find any frame based codec decoding? What's the difference between frame based and stream based encoding?
[03:26] <vl4kn0> decoding*
[09:28] <SirCmpwn> so I had an idea
[09:28] <SirCmpwn> I'd like to render subtitles out of a video, onto a transparent background
[09:29] <SirCmpwn> I tried ffmpeg -i transparent.png -video_size 1920x1080 -vf "subtitles=output.ass" -to 00:01:34 test.mp4
[09:29] <SirCmpwn> but the resulting file is unplayable
[09:43] <SirCmpwn> this works if transparent.png is already the right size: ffmpeg -y -loop 1 -i transparent.png -vf "scale=1920:1080" -c:v libx264 -pix_fmt yuv420p -to 00:01:34 -vf "ass=output.ass" test.mp4
[09:43] <SirCmpwn> so apparently -vf scale doesn't work on images?
[10:07] <anshul> it works at my place
[10:08] <anshul> can you try after  combining both filter  -vf "scale=1920:1080;ass=output.ass"
[10:09] <anshul> some time sccaling fails when out pix fmt is different
[10:10] <anshul> try first converting your image to yuv420p then do scaling, oviously it is an work around
[11:22] <Spideru> MOrning
[11:23] <Spideru> http://trac.ffmpeg.org/wiki is down
[12:18] <niakpi> pyffmpeg won't import: "DLL load failed: specified module cannot be found" (Win32, Python26)
[12:18] <niakpi> any suggestions?
[12:18] <JEEB> that thing probably has nothing to do with the FFmpeg project itself :)
[12:18] <JEEB> because as far as i know such wrapper isn't maintained by FFmpeg
[12:18] <JEEB> and given that python version I'm guessing that wrapper is just old
[12:19] <JEEB> there's been quite a few API changes during the years
[12:20] <niakpi> http://code.google.com/p/pyffmpeg/
[12:20] <niakpi> It's a different team I think
[12:20] <JEEB> yes, completely unrelated
[12:20] <JEEB> OUCH
[12:21] <JEEB> four years old
[12:21] <JEEB> yes
[12:21] <niakpi> I know
[12:21] <JEEB> that is most definitely going to have to be brought up-to-date
[12:21] <niakpi> you know any alternatives?
[12:21] <JEEB> because APIs have changed
[12:21] <JEEB> nope
[12:21] <sacarasc> The latest alpha on that site is almost 3 years old.
[12:21] <JEEB> https://github.com/tranx/pyffmpeg/branches
[12:21] <JEEB> their repo says 4 years since last commit
[12:21] <JEEB> :D
[12:21] <sacarasc> Wow.
[12:22] <niakpi> I need it for opencv VideoCapture
[12:22] <niakpi> in pytho
[12:22] <niakpi> 8python
[12:54] <fschuetz> Hello.
[12:55] <fschuetz> Does ffmpeg offer a way to play from a filedescriptor - for example from an incoming stream of data?
[12:55] <barhom> So I know I can use ffmpeg to transcode into h264+aac inside a mpegts container and stream the output to a multicast address and then playback via VLC. That is no issue. I would like to do the same with webm, vp8/vorbis. What am I missing? Doesn't webm support sending the metadata every x packets like mpegts to playback can start at any point?
[13:41] <Hello71> fschuetz: /dev/fd/0
[13:41] <Hello71> note that seeking may fail
[13:41] <Hello71> actually, just -
[13:42] <Hello71> some containers are not streamable, however
[13:42] <Hello71> e.g. mp4
[13:45] <Spideru> Hi, I'm trying to load this page https://trac.ffmpeg.org/ but from yesterday is down
[14:04] <Overdrive> Hi guys, so i searched the whole internet and i could not find a guide that matched ffmpeg.exe commands to C++ settings for x264 encoding settings. Do you know where I could find such a guide?
[14:05] <Overdrive> For exampe i've seen that it is advised to set options like this now: av_opt_set(c->priv_data, "preset", "slow", 0);. But what does slow mean, and what are the other options that I can set in that way? Is there a list somewhere ?
[14:06] <sacarasc> Overdrive: I'd point you to https://trac.ffmpeg.org/wiki/x264EncodingGuide, but it is currently down.
[14:06] <Overdrive> I've seen references on more websites to it but it is down for at least one week.
[14:22] <Overdrive> Ok so i managed to see the page using the wayback machine. I've seen references to flags2=+wpred+dct8x8. But in c there are no CODEC_FLAG2_ dct8x8 or anything simlar.
[17:23] <zmuss> Hi, I need help with creating rtsp stream with ffserver. There is no sound. I'm using aac codec. This is my ffservers.conf file: http://pastebin.kde.org/pxcqnpgup
[17:29] <marti____> hello?
[17:30] <marti____> somebody knows why https://trac.ffmpeg.org is offline?
[19:12] <juke> hi
[19:13] <juke> i convert a stream to many images, is there a way to limit dest directory size ?
[19:48] <juke> is there a way to manipulate ffmpeg via socket/telnet ?
[20:09] <DeadSix27> is possible via ffmpeg, to have 2 2channel audio input, and map 1 of them to FR/FL/C and the other 2 to SR and SL?
[20:32] <grkblood> im trying to set the chunk duration of my stream with `ffmpeg -i stream.mp3 -acodec libspeex -ar 8000 -ac 1 chunk_duration 10000 -f ogg pipe:1` and I'm getting the error Unable to find a suitable output format for 'chunk_duration'. how do I set this?
[20:46] <YuGiOhJCJ> hello do you understand why I got an "ALSA buffer xrun" message and an audio desync when I am capturing with ffmpeg (474db7a696a34582ee655c07e5073a6702236be0)?
[20:52] <ChocolateArmpits> grkblood, did you forget to put a dash before chunk_duration ?
[20:53] <grkblood> no, there is no dash, not according to the docs anyways.
[20:54] <grkblood> if I put a dash it doesnt recognize the command
[20:57] <ChocolateArmpits> ok, is your build compiled with speex?
[21:05] <grkblood> yep
[21:09] <grkblood> it works without the chunk_duration paramater ChocolateArmpits
[21:10] <grkblood> this will be used for live streaming so i need as little delay as possible and with the speex compression each chunk is about 5 seconds. so theres a 5 second delay
[21:10] <grkblood> i need to decrease the chunk size/duration
[21:12] <grkblood> and the wiki is down https://trac.ffmpeg.org/wiki/StreamingGuide#Latency
[21:15] <ChocolateArmpits> google cache has that page
[21:30] <saste> ubitux: ping
[21:30] <ubitux> saste: pong
[21:30] <saste> ubitux, same as yesterday
[21:31] <ubitux> yes, i'll answer in a sec on -devel
[21:31] <saste> let me know if I should remove the lavfi entry from the gsoc page
[22:39] <YuGiOhJCJ> http://pastebin.com/d3Sd69hm
[22:39] <llogan> YuGiOhJCJ: what player are you using?
[22:40] <YuGiOhJCJ> VLC
[22:41] <llogan> ffmpeg -y -f x11grab -framerate 15 -video_size 640x480 -i :0.0 -f alsa -i default -c:v libx264 -qp 0 -preset ultrafast -c:a copy out.mkv
[22:42] <llogan> if that doesn't work then try capturing video and audio using two separate ffmpeg processes and muxing the resulting files together
[22:42] <llogan> if it does work, then you can re-encode out.mkv (since it is lossless and therefore probably huge-ish)
[22:44] Action: llogan leaves to replace alternator in shitty car
[22:45] <YuGiOhJCJ> in fact it does not help but I found that the problem is that my current directory is an NFS filesystem that is slow
[22:45] <YuGiOhJCJ> if I go to the /tmp directory then I don't get anymore the "ALSA buffer xrun" message
[22:46] <YuGiOhJCJ> because it is now a normal ext4 filesystem (not NFS)
[22:46] <YuGiOhJCJ> and it is faster
[22:48] <YuGiOhJCJ> I got a similar problem when I am streaming on an external server (like twitch.tv)
[22:48] <YuGiOhJCJ> the "ALSA buffer xrun" message is back again
[22:49] <YuGiOhJCJ> and so my audio stream is out of sync
[22:50] <YuGiOhJCJ> for streaming I am not sure that it is possible to capture using two separate ffmpeg processes... So is there an other solution for streaming ?
[00:00] --- Sun Feb  9 2014