[Ffmpeg-devel-irc] ffmpeg.log.20120505

[burek]

[00:07] <TH0MAS_> Is it possible to use ffmpeg to stream on demand for use with Wowza
[00:07] <TH0MAS_> Ex. I want to stream an .AVI but Wowza does not have support..Can I somehow everytime the media is called on stream on the fly so that wowza can except it and play it
[00:10] <Mavrik_> that would probably be complicated to setup
[00:10] <Mavrik_> have you tried using wowza transcoder?
[00:17] <TH0MAS_> probably a better idea to just encode them offline
[00:18] <TH0MAS_> Is there a collection of arguments that work well for certain files
[00:18] <TH0MAS_> ex. avi to flv or avi to mp4
[00:27] <pasteeater> TH0MAS_: depends on the specifics of the output that you want and how you're going to use it
[00:27] <pasteeater> "flv" and "mp4" are container formats and can therefore contain various video formats
[00:28] <pasteeater> so you need to be more specific as to what you want
[00:28] <pasteeater> although I assume H.264 video.
[00:28] <TH0MAS_> well since I am using Wowza to stream my content Im going to have to convert
[00:28] <TH0MAS_> and keep it as close as a 1:1 as possible
[00:28] <TH0MAS_> and at the same time try to compress it enough to save bandwidth
[02:17] <desti> http://pastebin.com/5MQrQXJb can you see there anything wrong? i only got 5 fps, but the source is 20 fps and the encoder uses only 20% cpu time on one core (was using the same with earlier ffmpeg and it was working there some weeks ago)
[02:32] <TH0MAS_> av_interleaved_write_frame(): Operation not permitted6 bitrate=5502.8kbits/s
[02:32] <TH0MAS_> anyone know what this is from
[02:38] <_SKiTZO> is there some way to have avconv/ffmpeg function as a rudimentary server that another playback client (such as VLC) can connect to and stream from over the network?
[02:40] <_SKiTZO> I see that there are a lot of protocols such as rtsp mentioned in the documentation
[03:34] <_SKiTZO> ok i am now using -f rtp  rtp://127.0.0.1:8090 and it is apparently streaming away my file
[03:34] <_SKiTZO> but how can I access that strem from for example vlc?
[03:35] <_SKiTZO> i tried just typing in the ip:port and seleted rtp as protocol
[03:35] <_SKiTZO> I also saved out an SDP from ffmpeg, but couldn't find where to put that in vlc
[03:43] <_SKiTZO> i managed to open the SDP from vlc but the output is just a black frame. i guess i won't be able to use ffmpeg as a rudimentary streaming server after al :(
[06:02] <TH0MAS_> ffmpeg -i input.mkv -vcodec copy -acodec aac -sameq output.mp4
[06:03] <TH0MAS_> anyone can tell me why that yields: Error while opening encoder for output stream #0:1 - maybe incorrect parameters such as bit_rate, rate, width or height
[06:19] <TH0MAS_> -strict experimental and it worked
[06:20] <TH0MAS_> when ffmpeg is running does the time displayed mean the location of the file it is at?
[06:20] <TH0MAS_> in minutes
[08:59] Action: itrufeng how to play wma use ffmpeg.
[09:00] Action: itrufeng i need a demo.
[09:00] Action: itrufeng thanks.
[09:06] Action: itrufeng how to play wma use ffmpeg.i need a demo.help.
[09:18] Action: itrufeng hi.everyone
[11:10] <buhman> so I'm trying to make really tiny pcm audio for playback primarily on a microcontroller
[11:11] <buhman> I want to get this so that a 3 minute song is less than 100KB; I started with "-ar 8000 -ac 1 -acodec pcm_u8" but that was about 10x too large, so then I lowered the sample rate to 1000
[11:12] <buhman> this seems to be quite close to something I could easily deal with on my microcontroller, but it seems I can't playback that output with ffplay/mplayer/etc..
[11:13] <buhman> ffplay show me some bizzare black and white graph looking thing and plays some odd noise
[11:14] <buhman> mplayer just plays it at what sounds like roughly 8x (that is, playing 1kHz as if it was 8kHz
[11:23] <buhman> I've also observed that it's not until somewhere between 3kHz and 2kHz that audio quality /significantly/ degrades
[11:24] <buhman> somewhere between 2kHz and 1kHz ffplay's output becomes unrecognizable from the original audio
[11:25] <buhman> I'm not sure how to determine if that's an encoding bug, or a playback bug, or if the audio quality really just drops that sharply
[11:34] <buhman> http://absdev.org/ffplay-graph.png the top is the graph from 3kHz (and basically indistinguishable from the 44.1kHz source) in comparison to 1kHz output
[11:35] <buhman> it seems a significant amount of audio data completely vanishes
[11:39] <joeconway_> Hi, i was hoping someone might be able to offer me some help with regards to sending an h.264 stream over rtp. I've set up the RTP stream with avformat but I can't work out how to use SDP files as VLC won't accept the RTP connection without one. I've found the function that puts an SDP file into a buffer, do I have to just send that buffer over the RTP connection as the first packet?
[11:39] <joeconway_> I'd really appreciate any help
[11:40] <buhman> joeconway_: who wouldn't ? :P
[11:40] <joeconway_> :p I've avoided bothering the IRC for the last few months but my dissertation is in in a few days so desperate times call for desperate measures haha
[11:42] <buhman> joeconway_: I'd hardly call using irc "desperate"
[11:42] <joeconway_> I really don't like to bother other people with my coding problems as I'm sure people have better things to be doing
[11:43] <joeconway_> also if you ask for help all the time you don't really learn as much I've found
[11:43] <joeconway_> unless you really are stuck
[11:43] <buhman> I disagree; discussing things seems to catalyze learning
[11:47] <joeconway_> thats true, of course. But I've always felt as though it must be frustrating for an 'expert' to be approach by someone who hasn't tried to get to grips with the problem domain. I prefer to have informed discussion to avoid the 'do my homework for me please' impression, because nobody likes that
[11:47] <joeconway_> shifts the dynamic from helping someone, to doing their work for them i guess
[11:47] <joeconway_> perhaps I'm just too british and stubborn
[11:47] <joeconway_> prefer to fail silently
[11:47] <joeconway_> ha
[13:31] <lasher> sorry to be a noob. I've been at this a while now... .webm: Unknown format
[13:32] <lasher> i just wanna convert webm to mp4 or anything else actually usefull
[13:33] <lasher> FFmpeg version SVN-r0.5.1-4:0.5.1-1ubuntu1.3
[13:33] <JEEB> that is old
[13:34] <JEEB> older than the format you are trying to load
[13:34] <JEEB> (at least its publicizing)
[13:38] <lasher> so I need a newer ffmpeg? or codecs for it?
[13:38] <lasher> and do you know if Arista Transcoder is using ffmpeg?
[13:41] <JEEB> no idea, and yes -- you will have to compile or otherwise get a newer ffmpeg
[13:42] <lasher> ok thnx
[14:42] <lasher> ahah! Ijust needed libavcodec-extra-52 and it seems to be happy. thanx again JEEB
[16:27] <gusto> hey
[16:27] <gusto> i just tried out to encode a video using theora, i do not use it, i just wanted to try it, and there i see that it ignores my vquscale
[16:28] <gusto> also h264 by libx264 does it, but there it is at crf=23 what is fine, but i would like to change that to 25 someday too
[16:29] <gusto> encoding a video with libtheora gives me always q=0 even if i have -q:v 10 or whatver, it is always at 0 and it encodes a 10 MBit/s stream that is far more what the original is - about 2,5 Mbit/s
[16:29] <gusto> some ideas how to disable that "superior" quality setting?
[16:32] <gusto> however, i am very satisfied with the mjpeg codec in ffmpeg ... it encodes to about 6 mbit/s on -q:v 3 and that is a good number
[16:51] <francogrex> what are some parameters to reduce the size of this:
[16:51] <francogrex> ffmpeg -i movie.flv -s 400x240 -aspect 2:1 -r 20 -vcodec mjpeg -qscale 1 -acodec adpcm_ima_wav -ac 2 movie.avi
[16:52] <francogrex> I can reduce -r to 5 but that makes the movie choppy
[16:52] <juanmabc> -vb NUMBERk
[16:52] <francogrex> what?
[16:52] <juanmabc> video bit rate
[16:53] <francogrex> ah -b ?
[16:53] <juanmabc> as in HD normal, low
[16:53] <juanmabc> 800k would me middle quality
[16:53] <juanmabc> ffmpeg -i file.flv
[16:53] <juanmabc> check bit rate on video stream
[16:53] <francogrex> ok I think by default it's taking 1500k
[16:54] <francogrex> bitrate: 1118 kb/s
[16:54] <francogrex> ok, and r can be reduced as well? anything else?
[16:55] <juanmabc> r is frames per second, so in animated movies, not so much from 10
[16:55] <juanmabc> and still
[16:55] <francogrex> 10 is minimum then
[16:56] <francogrex> can I reduce audio quality? will it have an impact?
[16:56] <juanmabc> sure -ab
[16:56] <francogrex> and -qscale ?
[16:56] <francogrex> juanmabc: thanks
[16:56] <juanmabc> ok
[16:58] <francogrex> if I fix the bitrate i think I don't need qscale anymore right?
[16:59] <juanmabc> no idea, i don't use it
[16:59] <juanmabc> let's see :D
[17:02] <gusto> you can use both then it chooses whatever is near
[17:02] <gusto> but i would not use -vbitrate i would use -vmaxbitrate and -vminbitrate, and that last can be omitted by -vqscale
[17:03] <francogrex> so -vmaxbitrate 800k ?
[17:04] <francogrex> or rather -maxrate 800k
[17:06] <francogrex> doesn't accept  -maxrate 800k
[17:07] <francogrex> -maxrate 4000k -bufsize 1835k needs bufsize
[17:35] <lightburst> I finally was able to successfully compile a demo program but it seems to crash when av_malloc()/av_mallocz() is invoked. I don't know why.
[18:01] <TH0MAS_> looking for some help with a video I encoded
[18:02] <TH0MAS_> it was an mkv and I am moving it to mp4 for compatibility with Wowza Media Server
[18:02] <TH0MAS_> http://i49.tinypic.com/2ztlk4j.jpg
[18:02] <TH0MAS_> This is what the file shows in VLC and it plays in VLC
[18:03] <TH0MAS_> but when I try to stream it, it fails and I am wondering if my encode is not support by Wowza
[18:03] <TH0MAS_> I did a did a -vcodec copy and -acodec acc
[18:04] <juanmabc> and that wowza support h264?
[18:36] <burek> gusto did you try with -b (-b:v)
[18:37] <burek> to limit the bitrate instead of limiting the quality with -q
[18:39] <burek> lightburst, check the source code for av_malloc usage
[18:39] <burek> source code of ffmpeg*
[18:40] <burek> TH0MAS_, did you set global header?
[18:45] <gusto> burek: do you think that would be a good idea?
[18:46] <gusto> burek: well, -b:v works
[18:46] <burek> I'm just guessing that not all encoder have an option to control the output through some -q (quality) param
[18:46] <gusto> burek: when i limit it to 1000 it goes about 900 kbps
[18:46] <burek> but most of them have an option to control the output through bitrate
[18:47] <burek> read about maxbitrate
[18:47] <gusto> so that would be OK, but why does quantizer ...
[18:47] <gusto> however
[18:47] <burek> -maxrate*
[18:51] <gusto> the output is ugly, but under really bad conditions like 64kbps it shows quite usable picture
[18:52] <gusto> looks like theora is good for videophony ;-)
[18:52] <burek> 64k for video?
[18:52] <burek> are you sure you know what you are doing?
[18:53] <gusto> however ... tihs maxrate and minrate is ignored too
[18:54] <gusto> burek: i just tried that out and it was surprisingly good, when you think of only 64kbps
[18:55] <burek> well, play around with params and find the perfect combination :)
[18:56] <gusto> i did
[18:56] <gusto> just setting -b:v to 2000k is OK
[18:57] <gusto> no surprise
[19:05] <TH0MAS_> burek I did not set a global header
[19:06] <TH0MAS_> is that the issue?
[19:06] <burek> it usually is
[19:06] <burek> because file storage formats usually put info about streams at the end of the file
[19:06] <burek> and for streaming, you need that to be sent first
[19:07] <burek> -flags +global_header
[19:07] <TH0MAS_> thats the only argument?
[19:11] <burek> or you can always use your best friend Google :)
[19:17] <TH0MAS_> crossing my fingers here
[19:24] <TH0MAS_> ffmpeg -i big_buck_bunny_1080p_stereo.avi -s 1080x608 -y -strict experimental -acodec aac -ab 128k -ac 2 -ar 48000 -vcodec libx264 -vpre default -vpre main -r 24 -g 48 -b 1400000 -threads 64 bigbuckbunny_1500.mov
[19:24] <TH0MAS_> that works but the quality is horrible
[19:29] <gusto> well
[19:29] <gusto> i would put out that aac audio, and make it mp3 so you can get rid of the strict
[19:29] <gusto> then maybe some HQ options for mpeg4?
[19:30] <burek> or use libaacplus
[19:30] <gusto> -mbd rd -flags +mv4+aic -trellis 2 -cmp 2 -subcmp 2 -g 480
[19:30] <burek> to get hiQ audio at 32kbps
[19:30] <gusto> what should that -b option mean? get rid of it
[19:31] <gusto> and do you have 64 processors to use 64 threads?
[19:32] <cbreak> you don't need 64 CPUs
[19:32] <gusto> however, 64 threads seems too much to me anyway
[19:32] <gusto> make it the number of processors
[19:35] <cbreak> I'd go for about 50% more than the number of _cores_ at most
[19:35] <cbreak> with two CPUs here, I usually use about 8 to 12 threads for compute bound things
[19:36] <cbreak> the difference from 8 upwards (the physical core count) is not that big
[19:36] <cbreak> but for IO bound things it still makes a tiny bit
[19:39] <TH0MAS_> well the passthru video worked seemed i was giving VLC an bad rtmp link
[19:39] <TH0MAS_> still hvae to see if the audio i sok
[19:40] <TH0MAS_> yes the audio is very very distored
[19:40] <TH0MAS_> ffmpeg -i input.mkv -vcodec copy -acodec aac -sameq output.mp4
[19:40] <TH0MAS_> using this
[19:40] <TH0MAS_> what can be done for that?
[19:40] <burek> what do you want to do?
[19:41] <TH0MAS_> improve the audio quality
[19:42] <burek> compile libaacplus
[19:42] <burek> and try it
[19:42] <burek> btw, you can't improve audio quality
[19:42] <burek> you can only keep it the same
[19:42] <TH0MAS_> it jus tseems like the encoding is off
[19:42] <burek> or use audio editor and improve it
[19:42] <TH0MAS_> sounds very distored
[19:42] <TH0MAS_> distorted
[19:42] <burek> can you please use pastebin.com, to show your command line and its output?
[19:43] <gusto> w8, let's stick to that threads discussion
[19:44] <gusto> what is the difference between exact the number cpu=threads and threads>=cpus?
[19:44] <burek> ?
[19:44] <gusto> is is "better" because more threads use more ram, or what?
[19:44] <burek> do you know how threads work?
[19:44] <gusto> a little bit yes
[19:45] <burek> well you should read a little bit more about it
[19:45] <gusto> sO?
[19:45] <burek> it's not (usually) ram related
[19:45] <burek> it's mostly cpu related
[19:45] <gusto> ok, so what makes up the perfromance boost (of the cpu)?
[19:45] <burek> better cpu for example :)
[19:46] <gusto> we assume that the CPU has not multi-threading
[19:46] <gusto> no
[19:46] <burek> ?
[19:46] <gusto> why should then be more threads than the CPU's are a performance improvement?
[19:47] <burek> you need to read about how they work, and it will be very clear then
[19:48] <gusto> well, so you do not know ...
[19:48] <burek> I do, but you don't :)
[19:48] <cbreak> most modern CPUs have multiple cores
[19:49] <cbreak> and each of those cores can have two "threads" with hyperthreading
[19:49] <burek> here: http://en.wikipedia.org/wiki/Thread_(computer_science)#Multithreading
[19:49] <cbreak> note that those CPU threads are different from the threads you use in a program
[19:49] <cbreak> for example with a Dual CPU Xeon Mac Pro, you'd have 2 CPUs, 8 Cores and 16 Hardware Threads
[19:50] <cbreak> so you can run 16 programs at physically the same time.
[19:50] <burek> and here about cpu threads: http://en.wikipedia.org/wiki/Multithreading_(computer_architecture)
[19:53] <gusto> i said that we assume that the cpus do not have multithreading, or we can count that as an additional core per core
[19:54] <gusto> and i was asking why should the number of threads bigger than the count of cpus (optionally +multithreading)
[19:54] <gusto> in your example with the mac pro it would be cpus=threads in the meaning cpu=cores+multithreading
[19:54] <gusto> but my question was .. why more than that?
[19:55] <gusto> and ... finally .. i was not telling anything, i was asking
[19:55] <burek> it's always better if you have more servers then clients
[19:56] <burek> my mistake
[19:56] <burek> vice versa
[19:56] <burek> that way, if any client is, for some reason, not ready to continue with his process
[19:56] <burek> server can always switch to serve another free client
[19:56] <burek> so the cpu is always utilized
[19:56] <burek> that's why it's obviously better to have more threads than cpus
[19:57] <ffizz> hello. im new to momentovps and tried placing an order but provisioning always fails
[19:57] <ffizz> i have already deposited bitcoins
[19:57] <burek> ffizz, are you sure you're on the right channel?
[19:57] <ffizz> burek: no, wrong channel
[19:58] <cbreak> gusto: the number of CPUs is 2
[19:58] <cbreak> the number of Cores in that mac pro is 8
[19:58] <cbreak> so why only two threads?
[19:58] <cbreak> 8 threads would give you much higher performance (if your task is parallelizable)
[20:01] <TH0MAS_> http://pastebin.com/kPkxQEWs
[20:02] <TH0MAS_> here is the log for the distorted audio
[20:04] <burek> try using NeroAacEnc or libaacplus
[20:04] <burek> and see the difference in audio quality
[20:08] <TH0MAS_> and run the same command ?
[21:19] <gusto> so so, thats a better explanation
[21:19] <TH0MAS_> So I have figured out my audio issue
[21:19] <TH0MAS_> now I have a video one
[21:20] <TH0MAS_> ffmpeg -i output.mkv -flags +global_header -vcodec copy -acodec aac -ac 2 -ar 48000 -ab 192k -strict experimental input.mp4
[21:20] <TH0MAS_> is what I am using
[21:20] <TH0MAS_> at times the video gets pixelated
[21:21] <TH0MAS_> or distorted
[21:21] <TH0MAS_> Im not sure of the right termanology
[21:24] <gusto> >>>
[21:24] <gusto> hey
[21:24] <gusto> you are using -vcodec copy, so that should not touch the video
[21:25] <TH0MAS_> yes I know not too sure whats the issue then
[21:25] <TH0MAS_> its not the video is low quality
[21:25] <TH0MAS_> its that it gets distorted randomly
[21:26] <burek> then your original is damaged
[21:29] <TH0MAS_> its not the original the original plays fine
[21:29] <TH0MAS_> I also actually converted the video for testing
[21:30] <TH0MAS_> and that one plays fine
[21:30] <TH0MAS_> but the conversion takes about an hour
[21:30] <TH0MAS_> when the copy takes minutes
[21:30] <TH0MAS_> maybe I shall show you a video of it playing?
[21:31] <gusto> ah
[21:31] <gusto> so then it is clear
[21:31] <gusto> then you have a problem with the decoder
[21:32] <gusto> did you try to play that original with ffplay?
[21:32] <gusto> or, when you are using mkv ... why to hell dont you first export the video out of there?
[21:32] <gusto> and you can export the audio that way too
[21:33] <gusto> using mkvexport tracks output.mkv 1:video.??? 2: audio.???
[21:33] <burek> gusto, he is actually exporting it with -vcodec copy
[21:33] <gusto> i know, that's why i brought up my plan B
[21:33] <burek> the problem might be remuxing
[21:33] <burek> it happened to me almost always
[21:33] <gusto> and that is what i am talking about
[21:33] <burek> when converting mkv to mp4 with h264
[21:34] <burek> but I could never find a pattern
[21:34] <burek> to be able to submit a bug report
[21:34] <burek> so, if you are in a hurry, just reencode
[21:34] <gusto> yes, that is also an option, he must reencode in every way
[21:35] <gusto> i do reencode TV h264 streams
[21:35] <gusto> because their timing is broken
[21:37] <TH0MAS_> so i will not be able to use -vcodec copy
[21:37] <TH0MAS_> what should I use then
[21:39] <gusto> i would first try to get the video stream out using mkvexport and then see how it plays
[21:40] <TH0MAS_> i tried to grab it as quick as I an
[21:40] <TH0MAS_> can
[21:40] <TH0MAS_> http://youtu.be/OkCofSURE74
[21:43] <TH0MAS_> in the first 2 seconds id say there is the pixelation...it always corrects itself tho
[21:43] <TH0MAS_> now Ive done a 2pass on the same source video
[21:43] <TH0MAS_> ffmpeg -y -i INPUT -r 30000/1001 -b 2M -bt 4M -vcodec libx264 -pass 1 -vpre fastfirstpass -an output.mp4
[21:43] <TH0MAS_> ffmpeg -y -i INPUT -r 30000/1001 -b 2M -bt 4M -vcodec libx264 -pass 2 -vpre hq -acodec libfaac -ac 2 -ar 48000 -ab 192k output.mp4
[21:43] <TH0MAS_> and there is no pixelation like this
[21:44] <TH0MAS_> but the time it takes to encode it is very long but the quality is near perfect
[21:56] <sacarasc> Very old ffmpeg. :o
[21:57] <TH0MAS_> me?
[21:57] <sacarasc> Yeah.
[21:58] <TH0MAS_> thought I am on the latest
[21:58] <gusto> how do you know?
[21:58] <gusto> he did not print out the version no
[21:58] <sacarasc> -vpre hq hasn't been used in years.
[21:58] <TH0MAS_> ffmpeg version 0.10.2 Copyright (c) 2000-2012 the FFmpeg developers
[21:59] <TH0MAS_> oh thats just from a website "cheat sheet" but none the less it worked well
[21:59] <gusto> and what should be used now instead?
[22:00] <TH0MAS_> its just a preset
[22:00] <TH0MAS_> it doesnt come with those presets anymore
[22:00] <TH0MAS_> im thinking the pixel issue could be because of the bitrate or frame size..is that possible?
[22:00] <TH0MAS_> maybe its not a good size for streaming
[22:04] <TH0MAS_> http://pastebin.com/F8Vqxcyh
[22:04] <TH0MAS_> here is my log to the video that pixelates
[22:07] <burek> if you dont want to or dont know how to compile ffmpeg: http://ffmpeg.gusari.org/static/
[22:07] <burek> just download and run
[22:08] <TH0MAS_> i was installing thru rpm
[22:08] <TH0MAS_> I am running this on my CentOS server
[22:09] <burek> these are daily static builds
[22:09] <TH0MAS_> but does my version really have anything to do with this issue
[22:09] <burek> so you can always have the latest git
[22:09] <burek> for testing
[22:09] <burek> and you can always use rpm for stable binary
[22:10] <TH0MAS_> ok did you see the pixelation issue?
[22:10] <TH0MAS_> can it be due to bitrate?
[22:11] <burek> TH0MAS_, those pixelation artifacts
[22:12] <burek> are due to a non-present key frame at the beginning of the file
[22:12] <burek> so, all the delta frames (differences from the keyframe) that follow
[22:12] <burek> don't really have a basis to apply a diff to
[22:12] <burek> so you get video like that
[22:13] <burek> if you are cutting the video, cut it at the keyframe
[22:13] <burek> so you won't have any more such problems
[22:14] <TH0MAS_> i am not cutting it
[22:14] <TH0MAS_> just copying the video
[22:14] <burek> well, that is not entire video for sure
[22:14] <burek> someone did cut it
[22:15] <burek> and whoever cut it
[22:15] <burek> didn't cut it so that the first frame is a key frame
[22:16] <TH0MAS_> is this fixable ?
[22:16] <burek> do you have original video
[22:17] <TH0MAS_> the mkv is fixable
[22:17] <TH0MAS_> after encoding it..this issue is fixed
[22:17] <TH0MAS_> but like i said it takes a while to do that
[22:18] <burek> look
[22:18] <burek> if you are cutting the video
[22:18] <burek> cut it so that the output begins at key frame
[22:18] <burek> ok?
[22:19] <burek> or re-encode your video completely, so the encoder will generate the output correctly anyway
[22:19] <TH0MAS_> i am open to re-encode
[22:20] <TH0MAS_> but what would be the fastest possible while keep a near copy of the source
[22:20] <burek> http://forum.doom9.org/archive/index.php/t-38387.html
[22:31] <TH0MAS_> i am not cutting the video
[22:31] <TH0MAS_> and the video was made from bd disc
[23:04] <wrightcc> is there a way to split videos within some measure of precision? say 100 ms
[23:05] <burek> of course
[23:05] <burek> hh:mm:ss.mmm
[23:05] <burek> for -t and -ss
[23:05] <saste> wrightcc: check also the segment muxer, and this branch: http://gitorious.org/~saste/ffmpeg/sastes-ffmpeg/commits/misc-segment-fixes-20120206
[23:06] <saste> but the segment has some precision limits, since it always split video at key frames
[23:06] <wrightcc> yeah thats what im trying to deal with
[23:07] <wrightcc> it would be awesome if i could just say 'insert keyframe at xyz'
[23:08] <saste> wrightcc: you can, -force_key_frames
[23:09] <wrightcc> saste, so this will give me more precision?
[23:10] <saste> but keep in mind that the setting of the key frame is somehow unprecise, since you don't know the exact timestamp of each frame
[23:10] <saste> that's why in my branch I have the delta segment option
[23:11] <wrightcc> ok ill check first if i can get away with the normal setup
[23:12] <saste> with delta = (1/framerate)/2 you should be usually safe
[23:25] <sgfgdf> hi guys! i tried to cut some pars of an audio file by using ffmpeg -i in.m4a -acodec copy -ss 10 out.m4a and then when i do ffmpeg -i out.m4a somewhere is says "Duration: 00:03:23.24, start: 0.007007", when i try to examine other audio files they all start at zero. so i did ffmpeg -i out.m4a -acodec copy -ss 0 out2.m4a. is that stupid step and do i loose quality while doing such cuts using the -acodec copy option?
[23:28] <burek> you don't use quality
[23:29] <burek> and that's probably a bug in muxer
[23:29] <burek> if you can, please report it to the ffmpeg's bug trac
[23:29] <burek> use -> lose
[23:29] <burek> I'm too tired to type anymore..
[23:30] <sgfgdf> burek, what is wrong the number 7?
[23:32] <sgfgdf> burek, or that it shows start different from zero?
[23:33] <burek> I'm not sure, but doing -acodec copy, it copies buffer "as is"
[23:33] <burek> so, I can only guess it keeps time stamps also "as is"
[23:34] <burek> so it shows 0.007007 instead of 0
[23:34] <burek> you could check in the source code to make sure
[23:38] <sgfgdf> burek, this information is just metadata or have something to do with playing the audio file? because when i did the last step trying to reset the "start" to zero then i did a diff between out.m4a and out2.m4a -- they differ. i hope the difference is only in this metadata not in the streams.
[23:40] <burek> sgfgdf, they differ in container only
[23:40] <burek> and that's good step
[23:40] <burek> to "reset" the start
[23:41] <burek> I'm afraid there is a bug in ffmpeg that prevents it from correctly writing the output in the first command
[23:41] <burek> and if you could report it, we could fix it :)
[23:42] <sgfgdf> burek, so the normal behaviour will be when for example i cut 10 seconds from file a into file b and then examine file b it should show start as zero, is that correct?
[23:44] <burek> yes
[23:44] <burek> timestamps should always start from zero
[23:44] <burek> or some same value (if both video/audio is present in the container)
[23:44] <burek> to be in sync
[23:46] <sgfgdf> burek, okay, thank you! will try to report this.
[23:49] <burek> thank you :)
[23:51] <sgfgdf> btw if you remember we talked about aac+ these days and you told me i can use libaacplus to reencode mp3s to aac+. i tried to reencode some mp3s with your suggestion using -ab 32k. in my case it appears that the new file is worse than the old one so i changed it to 64, but i coudn't get higher.
[23:52] <sgfgdf> is this the maximum bit rate i can get with this library?
[23:52] <burek> worse?
[23:52] <burek> that's odd
[23:52] <burek> do you have cmd line + output
[23:52] <burek> so I can examine?
[23:52] <sgfgdf> with 32kbps, but 64 seems to be good.
[00:00] --- Sun May  6 2012