[Ffmpeg-devel-irc] ffmpeg.log.20121105

[burek]

[00:00] <sean1988> im getting compile errors
[00:00] <darkstarbyte_> This sounds bad.
[00:00] <darkstarbyte_> what are they?
[00:00] <sean1988> lots lol
[00:01] <darkstarbyte_> warnings or errors?
[00:01] <sean1988> warnings
[00:01] <darkstarbyte_> It could be because the strict setting they always set
[00:02] <darkstarbyte_> Usually it is because you don't include every possible dependency that it can have.
[00:02] <sean1988> ah right well its installed so i'll have a try
[00:02] <darkstarbyte_> The whole script finished?
[00:03] <sean1988> yup
[00:03] <darkstarbyte_> That is a fast machine you have there.
[00:03] <sean1988> over clocked to 4.4GHz
[00:03] <sean1988> 16gb ram
[00:03] <darkstarbyte_> That is why
[00:03] <darkstarbyte_> Double the cores at more than twice the clock of my machine.
[00:03] <sean1988> yeah i shouldn't be getting 120 fps when converting and avi to mpg
[00:04] <darkstarbyte_> I get 600fps encoding mpeg2
[00:04] <sean1988> see my point?
[00:04] <sean1988> maybe I should try gentoo
[00:04] <darkstarbyte_> And that is using only half my cpu power which would be technically 800Mhz across 4 cores
[00:05] <darkstarbyte_> wait
[00:05] <darkstarbyte_> I have one more idea
[00:05] <sean1988> hold up i ain't tried this yet lol
[00:05] <darkstarbyte_> ffmpeg -threads 8 -i input.ext  -acodec acodec -vcodec vcodec -threads 8 output.ext
[00:06] <sean1988> im guessing with file names to
[00:07] <darkstarbyte_> I could just write the line up
[00:07] <sean1988> i think somthing failed to install or compile
[00:07] <darkstarbyte_> hmm?
[00:07] <darkstarbyte_> which one?
[00:07] <sean1988> Unknown encoder 'vcodec'
[00:07] <darkstarbyte_> oh
[00:07] <darkstarbyte_> I did not put in the codec names
[00:07] <darkstarbyte_> I will do that now
[00:08] <sean1988> I jsut wanna say thank you for your help this has been annoying me for so long now.
[00:08] <darkstarbyte_> ffmpeg -threads 8 -i input.ext  -acodec ac3 -vcodec mpeg2video -b:v 16000k -b:a 448k -threads 8 output.ext
[00:08] <darkstarbyte_> Your welcome
[00:09] <darkstarbyte_> That should be a high quality video
[00:09] <darkstarbyte_> , because 1080p uses around 22000k bitrate
[00:09] <sean1988> getting 170 fps
[00:10] <darkstarbyte_> wow
[00:10] <darkstarbyte_> Can you pull up a
[00:10] <darkstarbyte_> task manager
[00:10] <sean1988> yup
[00:10] <darkstarbyte_> It should show you how much your cpu is being stressed.
[00:11] <sean1988> 50% on 2 at max
[00:11] <darkstarbyte_> Your computer is bottlenecking like mine.
[00:11] <darkstarbyte_> Lets try more threads
[00:11] <darkstarbyte_> ffmpeg -threads 16 -i input.ext  -acodec ac3 -vcodec mpeg2video -b:v 16000k -b:a 448k -threads 16 output.ext
[00:12] <sean1988> started off higher but now it's drped to 150
[00:12] <sean1988> but only started at 200
[00:12] <darkstarbyte_> Lets try x262
[00:13] <darkstarbyte_> git clone https://github.com/kierank/x262
[00:13] <darkstarbyte_> cd x262
[00:13] <darkstarbyte_> ./configure
[00:13] <darkstarbyte_> make -j8
[00:13] <darkstarbyte_> then
[00:14] <darkstarbyte_> ./x262 -o output input
[00:14] <darkstarbyte_> Don't install it.
[00:14] <darkstarbyte_> Just want to test.
[00:15] <darkstarbyte_> sean1988, Any reason your using mpeg2 instead of something like h.264?
[00:15] <sean1988> to put them onto a dvd disc
[00:15] <darkstarbyte_> You must be using devede and decided to use the advanced options and have the video preconverted?
[00:17] <darkstarbyte_> x262 will at least use all of the cpu.
[00:17] <darkstarbyte_> I am not sure how much faster it will be though.
[00:18] <sean1988> Lol no, i was using a program in windows called convertxtodvd which uses ffmpeg in the back and when i went to linux i tried alot of diffrent ones and everyone said use ffmpeg on the comand line so i tried it but it was just as slow and then i found out the the windows software uses ffmpeg as well but runs much faster.
[00:18] <darkstarbyte_> See if that x262 helps.
[00:19] <sean1988> what is the extension on the output or does is not matteR?
[00:19] <darkstarbyte_> .mpeg
[00:19] <darkstarbyte_> it does
[00:19] <darkstarbyte_> wait one more thing
[00:19] <darkstarbyte_> use the option --mpeg2
[00:20] <sean1988> it can't open the avi
[00:21] <darkstarbyte_> OMG
[00:22] <darkstarbyte_> It should be able to.
[00:22] <darkstarbyte_> Mine can and I don't have ffmpegsource installed
[00:22] <sean1988> i did get alot og warnings when compiling
[00:23] <darkstarbyte_> This is really annoying.
[00:23] <darkstarbyte_> I see why were bothered by this.
[00:23] <sean1988> yup
[00:23] <darkstarbyte_> Some dvd players will play avi directly you know.
[00:24] <sean1988> some but not my mam and dads lol and thats a blu ray player
[00:24] <darkstarbyte_> woh woh woh
[00:24] <darkstarbyte_> a blu ray player that won't?
[00:24] <sean1988> yeah, but tbf it's rather old now lol
[00:25] <darkstarbyte_> You want to see something funny?
[00:25] <sean1988> sure
[00:26] <darkstarbyte_> http://www.linuxquestions.org/questions/linux-software-2/ffmpeg-mpeg2-encoding-bottleneck-and-i-cant-find-a-work-around-4175435204/
[00:26] <darkstarbyte_> That is my bottleneck thing I posted.
[00:26] <darkstarbyte_> I tested it really well.
[00:26] <darkstarbyte_> ram disk, overclocking, and I showed my work.
[00:28] <darkstarbyte_> I have to go make something, sorry I could not help you.
[00:28] <sean1988> you get faster speeds on a single core
[00:28] <darkstarbyte_> 3.06Ghz
[00:29] <darkstarbyte_> It is kind of funny now that I think of it.
[00:29] <sean1988> so even if it was using one one corretly i should get at least 220 right?
[00:29] <darkstarbyte_> yeah
[00:29] <darkstarbyte_> Some video cards get 1k+ fps
[00:29] <sean1988> Lol
[00:30] <darkstarbyte_> The ones with 100's of cores clocked at 200Mhz
[00:30] <sean1988> so what you hink i should do learn to install gentoo lol
[00:30] <darkstarbyte_> They also have faster ways of doing many operations.
[00:31] <darkstarbyte_> arch might be better
[00:31] <darkstarbyte_> Considering I am using Slackware
[00:31] <sean1988> yeah but at least i got arch to boot
[00:31] <darkstarbyte_> Arch is usually a good choice, and you can make your kernel custom for your computer
[00:31] <sean1988> although i couldn't get xorg to work right
[00:32] <darkstarbyte_> which can speed encoding 5% with a custom kernel alone.
[00:32] <darkstarbyte_> sean1988, I had to install my video special.
[00:32] <sean1988> I've tried a bunch of diffrent arch based distro but they wern't any good and couldn't get the graphic drivers to work right
[00:32] <darkstarbyte_> I have to go, so you around.
[00:33] <sean1988> you to mate
[00:33] <sean1988> i'll probly be here till i get this sorted
[00:33] <Mavrik> darkstarbyte_: btw, a small detail
[00:33] <Mavrik> in ffmpeg it matters where you put a parameter
[00:33] <Mavrik> ffmpeg -threads 8 -i ... means "8 threads for DECODER"
[00:34] <Mavrik> ffmpeg -i -codec:v mpeg2video -threads 8 -codec:a ... means "8 threads for MPEG-2 VIDEO encoder"
[00:34] <cbsrobot> dmon: try -vf mp=eq2
[00:34] <Mavrik> ffmpeg -i -codec:v mpeg2video -codec:a mp2 -threads 8 ... means "8 threads for MPEG-2 AUDIO encoder"
[00:34] <Mavrik> darkstarbyte_: so "ffmpeg -threads 8 -i ..." doesn't do what you think it does :)
[00:37] <sean1988> yeah still didn't help :(
[00:49] <sean1988> anyone else here want to try help me out?
[00:54] <darkstarbyte_> I do know of one thing that might help
[00:55] <sean1988> Yeah...
[00:55] <darkstarbyte_> strace -c ffmpeg (commands here)
[00:55] <darkstarbyte_> Not in parentheticals though.
[00:56] <sean1988> nope still no luck
[00:56] <darkstarbyte_> That is for me to debug something I think it could be.
[00:56] <sean1988> ah so should i keep it running till it's complete?
[00:57] <darkstarbyte_> yeah
[00:58] <darkstarbyte_> It will give you stuff at the end that could show if it is the kernel that is at fault
[00:58] <darkstarbyte_> Perhaps ubuntu did something stupid.
[00:58] <darkstarbyte_> or it is glibc
[00:59] <sean1988> i had the same problem with debian.
[00:59] <sean1988> I didn't try anything else because of graphic drivers
[00:59] <darkstarbyte_> This is getting strange
[01:00] <sean1988> i've droped to 80 fps
[01:00] <darkstarbyte_> Just need the output at the end
[01:01] <darkstarbyte_> I think you can now hit
[01:01] <darkstarbyte_> ctrl + c
[01:01] <sean1988> and what do you want pasted?
[01:02] <darkstarbyte_> It should just be a bunch of percentages.
[01:02] <darkstarbyte_> and time
[01:02] <sean1988> yeah should i just post them in here?
[01:02] <darkstarbyte_> no
[01:02] <darkstarbyte_> I can do a private chat to do it though.
[01:02] <sean1988> okay
[01:03] <darkstarbyte_> I opened it.
[01:03] <darkstarbyte_> You should see it.
[01:03] <sean1988> nope lol
[01:03] <darkstarbyte_> darn
[01:03] <darkstarbyte_> It should be next to the channels
[01:03] <sean1988> nothing there
[01:04] <darkstarbyte_> try dpaste.com
[01:04] <sean1988> will do
[01:04] <sean1988> http://dpaste.com/824652/
[01:04] <darkstarbyte_> wow
[01:05] <darkstarbyte_> something is wrong
[01:05] <darkstarbyte_> Most likely with ffmpeg
[01:05] <darkstarbyte_> That encoder has a major bottle neck.
[01:06] <sean1988> any ideas what it would be??
[01:06] <darkstarbyte_> Maverik had one, but he logged off before he could tell me.
[01:07] <sean1988> Lol.
[01:08] <sean1988> yeah also when i use just ffmpeg in windows it keeps coming up saying buffer underflow a bunch of times
[01:09] <darkstarbyte_> Perhaps there is another mpeg2 library that we can use, I will look it up.
[01:10] <sean1988> cool man
[01:14] <darkstarbyte_> mpeg2enc
[01:14] <darkstarbyte_> I will try to test it.
[01:15] <sean1988> okay bud
[01:16] <sean1988> i was thinking about trying somthing like that but I didn't really have a clue what or how to even start
[01:23] <darkstarbyte_> compiling mjpeg tools
[01:27] <darkstarbyte_> This is starting to get annoying.
[01:30] <sean1988> I'm gunna go for a smoke make i'll be back in 10
[01:33] <darkstarbyte_> I am trying to find how to write the output of ffmpeg to stdout
[01:48] <sean1988> i'm back
[01:48] <darkstarbyte_> I still can't find how to write to stdout
[01:49] <darkstarbyte_> with ffmpeg
[01:49] <sean1988> can i not just > pipe it to a .txt?
[01:49] <darkstarbyte_> found it
[01:49] <darkstarbyte_> yeah
[01:49] <sean1988> lol i do it all the time for manuals and stuff
[01:50] <darkstarbyte_> This is what I have
[01:50] <darkstarbyte_> ffmpeg -i dex.flv -vcodec huffyuv -f raw -an /dev/stdout | mpeg2enc -b 15000k -M 8 --level high -o dex.mpg
[01:50] <darkstarbyte_> Can't write to stdout
[01:50] <darkstarbyte_> without the /dev/stdout part
[01:50] <darkstarbyte_> oh
[01:50] <darkstarbyte_> pipe:
[01:52] <darkstarbyte_> Now I need a the raw output formate
[01:52] <darkstarbyte_> format
[01:53] <darkstarbyte_> wow
[01:53] <darkstarbyte_> they have a yuv4mpeg2pipe
[01:54] <darkstarbyte_> AAAHHHHHH
[01:56] <sean1988> I dunno what that means lol
[01:56] <darkstarbyte_> I think I will just write the raw file to my hard drive and go from there
[01:58] <darkstarbyte_> I am give up
[01:58] <darkstarbyte_> Someone else has got to know.
[02:09] <darkstarbyte_> night night, and I hope someone will help.
[06:06] <jordan_> Hello
[06:07] <hendry> is there an ffmpeg audio filter to normalise the volume of a video clip?
[06:24] <jordanreiter> Hello, is anyone else here?
[06:25] <jordanreiter> Hoping someone here can help me with a perplexing problem...
[06:25] <jordanreiter> I'd ask the users mailing list but sent a request to join and no response.
[06:26] <burek> hendry yes
[06:26] <burek> http://ffmpeg.org/ffmpeg.html#volume
[06:27] <burek> but im not sure how exactly to get it to do normalizing
[06:27] <jordanreiter> burek, would you be able to help me with a problem w/ffmpeg?
[06:27] <burek> maybe you should combine it with "volumedetect" filter
[06:27] <burek> jordanreiter, don't ask to ask, just ask :)
[06:28] <jordanreiter> Okay!
[06:28] <jordanreiter> Didn't want to barge in :)
[06:28] <jordanreiter> Here's the weirdness:
[06:28] <jordanreiter> I'm combining a PNG & a WAV file using the following command:
[06:29] <jordanreiter> ffmpeg -loop 1 -i input.png -i input.wav -shortest -y output.mpg
[06:30] <burek> really weird :)
[06:30] <jordanreiter> Now around 3/4 of the time it correctly generates a file of ~1.2MB, with these stats:
[06:30] <jordanreiter> frame=  302 fps= 84 q=31.0 Lsize=     594kB time=00:00:12.04 bitrate= 404.2kbits/s
[06:30] <jordanreiter> video:401kB audio:188kB subtitle:0 global headers:0kB muxing overhead 0.829003%
[06:30] <jordanreiter> BUT around 1/4 of the time, I get a teeny tiny file and an output that looks like this:
[06:31] <jordanreiter> frame=   24 fps=0.0 q=31.0 Lsize=     230kB time=00:00:11.99 bitrate= 157.1kbits/s
[06:31] <jordanreiter> video:39kB audio:188kB subtitle:0 global headers:0kB muxing overhead 1.643419%
[06:31] <jordanreiter> If I do a debug run, when it generates the file correctly there are a bunch of lines beginning with AVIOContext
[06:32] <jordanreiter> But when it generates the short file it skips straight from "Press [q] to stop, [?] for help" to the EOF
[06:32] <burek> does your output play ok?
[06:32] <burek> also, please use pastebin
[06:32] <burek> showing only 1-2 lines doesn't help at all
[06:33] <jordanreiter> It actually does, *until* I concat it with another mpg file. Then it plays the audio for both movies, but jumps straight from the first image to the second image after a few ms.
[06:33] <burek> provide a complete output if you want us to be able to see what's wrong
[06:33] <burek> how do you concat videos exactly?
[06:33] <jordanreiter> burek: I've done a diff between the two outputs. Literally the only difference is that when it *does* work correctly, the AVIOContext lines show up and when the file doesn't generate correctly, they don't.
[06:34] <jordanreiter> cat video1.mpg video2.mpg > combined.mpg
[06:34] <jordanreiter> then just do ffmpeg -i combined.mpg -qscale:v 2 final.mpg
[06:34] <burek> are you sure cat is not creating the problem?
[06:35] <jordanreiter> Nope, if I concat them when the correct file was generated, it works; if I concat them when the short file was generated, it doesn't.
[06:35] <jordanreiter> So the difference is when the initial generation code doesn't work.
[06:35] <burek> wait
[06:35] <jordanreiter> I'l go ahead and run through it, get some output, and put it in pastebin. Give me a minute?
[06:35] <burek> don't go with cat
[06:35] <burek> if you didn't test the output of ffmpeg first
[06:35] <burek> so, try to play the short file
[06:35] <burek> to see if it plays
[06:36] <jordanreiter> burek, it does play and appears to play correctly, either version.
[06:36] <jordanreiter> However, one version is much larger than the other. My guess is that the player compensates for the fact that it only has 3 frames and just keeps the image up there.
[06:36] <jordanreiter> But when the two video files are combined, because it's only say 8 frames, it jumps to the frames for the second movie.
[06:36] <jordanreiter> I'll get you some pastebin output.
[06:42] <jordanreiter> Okay, here's the result when the generation is "bad" http://pastebin.com/6U8k5zwD
[06:43] <jordanreiter> And here's the result when the generation is "good" http://pastebin.com/8s74ExzQ
[06:44] <jordanreiter> The only differences are: the memory addresses (0x279.... vs 0x777....) and the "Statistics" line followed by the second grouping of AVIOContext.
[06:44] <jordanreiter> Thanks for your help burek!
[06:45] <burek> [wav @ 0x27ac640] parser not found for codec pcm_s16le, packets or times may be invalid.
[06:45] <jordanreiter> I did an ls and the bad file is just 264k while the good file is 934k. I'm guessing the bad file is missing information in the video stream which is why it doesn't work correctly when concatenating.
[06:46] <jordanreiter> Hmmm. Yeah, I see that for both files; why would it work sometimes but not always? And where can I get the parser for that codec and/or how can I direct it to use a different codec?
[06:47] <jordanreiter> So if that is the problem, it looks like unfortunately it is intermittent and unpredictable. :(
[06:48] <burek> you can create a bug ticket if you want
[06:49] <jordanreiter> So is this definitely a bug? Is there any way for me to get this to work for now?
[06:50] <jordanreiter> I mean, is there any way to force ffmpeg to use the full length of the audio file to determine the number of frames?
[06:51] <jordanreiter> Hmmm... I'm using the trunk version. I'll try doing a pull, make it again, and see if that helps... *fingers crossed*
[06:58] <burek> well, try just ffplay
[06:58] <burek> for ex. ffplay input.wv
[06:58] <burek> wav*
[06:59] <burek> do it several times and if it plays ok, at least you'll be sure the input is not damaged or something
[07:02] <jordanreiter> Okay, I'll do that once it's recompiled. I'm hoping it was just a problem with the trunk version I was using.
[07:12] <jordanreiter> Hey back from recompiling. I don't have ffplay on the machine (and it's a remote server so I'm not sure whether I'd be able to get it to work)> I can definitely download it and try running it on my local computer.
[07:19] <jordanreiter> Okay, downloading ffplay to my local machine...
[07:27] <jordanreiter> burek, I don't think it's the wav file.
[07:39] <jordanreiter> Okay burek, thanks for your help so far. I've gotta go it's late here.
[11:03] <tuxx_> hey guys
[11:04] <tuxx_> i am struggling to fix a program called vnc2mpg... the program is an example source from libvncserver, which allows recording a vnc session and trasncoding it to mpeg using ffmpegs libavformat libraries.
[11:05] <tuxx_> however strangely when playing back a recorded a vnc session, the video plays way to fast.. i think that the pts (?) information is broken
[11:05] <tuxx_> from what i can see in the source video_pts is set but never used...
[11:06] <tuxx_> could someone point me to an example showing how time stamp information is correctly encoded into mpeg streams?
[12:33] <Yulth> Hi everyone!
[12:40] <Yulth> Is there any way to convert from medium-quality MP3 (128 - 192kbps) to 40 - 64kbps HE-AACv2 with no quality loss? (or with the minimum loss)
[12:41] <durandal_1707> Yulth: i dont think there is
[12:43] <Yulth> the target is obvious: to save bandwidth maintaining quality :)
[12:44] <durandal_1707> transcoding conflict with quality
[12:44] <darkstarbyte_> I would say with faac use 32Kb/s per channel
[12:44] <darkstarbyte_> if not more
[12:45] <darkstarbyte_> 6 channels = 192
[12:45] <darkstarbyte_> kbs
[12:45] <darkstarbyte_> minimum
[12:45] <divVerent> darkstarbyte_: rather more in case it's 2 channels :P
[12:46] <divVerent> I got at least 128kbit/s for faac to get acceptable quality for 2 channels
[12:46] <darkstarbyte_> yeah
[12:46] <divVerent> and it's still lower than LAME 128kbit/s MP3
[12:46] <darkstarbyte_> using faac?
[12:46] <divVerent> yes
[12:46] <darkstarbyte_> Something is up.
[12:47] <divVerent> yes, faac sucks
[12:47] <divVerent> same with commandline faac encoder
[12:47] <divVerent> I recommend trying libfdk_aac
[12:47] <ubitux> Yulth: transcode from lossless, lossy-to-lossy is evil and you will loose quality
[12:47] <darkstarbyte_> Perhaps celt or something then.
[12:47] <divVerent> although this one sometimes crashes
[12:47] <divVerent> darkstarbyte_: celt doesn't help IIRC
[12:47] <darkstarbyte_> I forgot what it is now called.
[12:47] <divVerent> if you want AAC in the end
[12:47] <darkstarbyte_> oh
[12:47] <darkstarbyte_> no flexibility?
[12:48] <divVerent> in my personal case, I want to playback on the iPhone
[12:48] <divVerent> no flexibility there
[12:48] <divVerent> can't even use libmp3lame
[12:48] <divVerent> Yulth may need AAC for another reason though
[12:48] <darkstarbyte_> android would not be so bad if it did not use java for nearly everything.
[12:48] <divVerent> I'd suggest trying libfdk_aac
[12:48] <divVerent> it's the best supported AAC encoder in ffmpeg
[12:48] <divVerent> but apparently somewhat buggy
[12:48] <divVerent> often crashes with assertion failed
[12:49] <divVerent> inside libfdk_aac codce
[12:49] <divVerent> from looking at the source, this can only happen at high volume levels and is roundoff error caused
[12:49] <divVerent> so, reduce volume by 1% if it happens :P
[12:50] <divVerent> while faac is rock solid stable
[12:51] <divVerent> oh, and by all means, do not use the ffmpeg-internal aac codec
[12:51] <divVerent> it is worse than faac :P
[12:56] <Yulth> ubitux: and is there any option to reach this target? (convert mp3 to a lower bitrate codec) or I'm damned with my entire mp3 collection? T_T
[12:56] <darkstarbyte_> T_T means your high.
[12:57] <ubitux> you can do lossy to lossy compression, but of course you're doomed with a mp3 collection...
[12:57] <Yulth> divVerent: save bandwidth. It's my main target
[12:57] <Yulth> frustrating.............. :S
[12:57] <ubitux> having lossless files will allow you to have a lossy copy to the best codec regularly
[12:57] <ubitux> lossy files are the end of the chain
[12:58] <divVerent> Yulth: your mp3s probably are lossy enough as is
[12:58] <divVerent> if you encode them again - ESPECIALLY when encoding using another codec than the original one - you WILL incure extra loss
[12:58] <divVerent> even if you encode to huge output bandwidths
[12:58] <divVerent> the question is rather, do you care
[12:58] <divVerent> I suggest you just try it out
[12:58] <divVerent> how far you get using the libfdk_aac encoder
[12:59] <divVerent> if you don't notice a quality loss, it's good for you
[12:59] <divVerent> in theory though, there is always a loss when transcoding between different lossy codecs
[12:59] <divVerent> and POTENTIALLY no loss when transcoding to the same codec and same or higher bitrate (but usually even this is lossy)
[13:00] <divVerent> thing is, AAC and MP3 may work differently enough so that transcoding mp3 128k to 96k may even yield better quality than mp3 128k to aac 96k
[13:00] <divVerent> for vorbis input, I'd actually BET on that ;)
[13:00] <divVerent> I mean, mp3 128k to mp3 96k is likely to sound better than mp3 128k to vorbis 96k
[13:00] <divVerent> despite vorbis performing a lot better at low bitrates
[13:01] <divVerent> I'd even bet on that for Opus, actually
[13:01] <Yulth> I understand......
[13:01] <divVerent> the issue is that different codecs cause different kind of loss, and you'll never get back what e.g. MP3 already cut off
[13:01] <divVerent> when you transcode
[13:02] <divVerent> still - try it
[13:03] <Yulth> I've made several tests converting 128 and 192kbps mp3 to 40-64kbps HE-AACv2 using libfdk_aac and I can tell you that the quality loss is terrible!!!
[13:06] <divVerent> well, can't expect much from 40 to 64kbps anyway
[13:06] <divVerent> even with AAC
[13:06] <divVerent> I'd expect aac 64kbps to be roughly like mp3 128kbps
[13:06] <darkstarbyte_> With celt dual channel is supposed to sound ok at 64kbps
[13:06] <divVerent> but causing different losses, so mp3 128kbps to aac 64kbps kinda combines the artifacts of both
[13:07] <divVerent> darkstarbyte_: same with aac, "ok"
[13:07] <divVerent> i.e. like 128kbps mp3 :P
[13:08] <Yulth> darkstarbyte_: what is "celt"?
[13:08] <divVerent> another codec
[13:08] <divVerent> the one that became the base of opus
[13:09] <darkstarbyte_> yeah
[13:09] <darkstarbyte_> Though I should have said opus instead.
[13:10] <divVerent> opus is good because it seems like it's actually there to stay
[13:10] <darkstarbyte_> In about 3 years from now I will make a large donation to xiph
[13:10] <divVerent> vorbis is quite good already, also regarding application support
[13:10] <Yulth> Supposedly, libfdk_aac is the best open-source HE-AACv2 converter and it's be able to produce professional aac quality. How it's possible that HE-AACv2 at 64kbps be like mp3 at 128kpbs? As far as I know, HE-AACv2 at 64kbps is the standard for High Quality Audio
[13:10] <divVerent> and I'd soon expect the same or better level of support for opus
[13:11] <divVerent> Yulth: HAHA
[13:11] <divVerent> nobody really calls 64kbps high quality audio
[13:11] <darkstarbyte_> It will start with movie pirates, I bet you that.
[13:11] <Yulth> :)
[13:11] <divVerent> in fact, I'd bet that you just cannot achieve "very good" quality with 64kbps whatever you do
[13:11] <divVerent> we'll soon hit the limit of audio compressability
[13:11] <divVerent> 32kbps is right out
[13:12] <darkstarbyte_> 32kbps is the real goal here
[13:12] <divVerent> 64kbps, maybe transparent to most people, but sure won't pass all ABX tests
[13:12] <divVerent> in the end, to find a lower limit of audio bitrate
[13:12] <darkstarbyte_> Great sound at 32kbps is what I would really love to see.
[13:12] <divVerent> how much entropy does the brain retain when processing audio input from the ears
[13:12] <divVerent> you can't go lower than that
[13:13] <divVerent> and I estimate that somewhere around 16kbps to 64kbps, depending on person
[13:13] <divVerent> just given how development of audio codecs seems to be
[13:14] <divVerent> like, look at a conductor of an orchestra
[13:14] <tuxx_> hey guys... can someone help me with vnc2mpg... the program seems very small and straightforward
[13:14] <divVerent> he can focus on anyone playing
[13:14] <divVerent> despite only having two ears
[13:14] <Yulth> All majors online radio say that AAC at 64 kbps is Very High Quality. Are they lying?
[13:14] <darkstarbyte_> The pirates might as well use h.264 on the highest settings with opus audio in the mkv format, then compress with 7z -42 and AES
[13:14] <divVerent> Yulth: yes :P
[13:14] <divVerent> or rather
[13:15] <tuxx_> but for some reason the video files it generates have no time information.. duration is 0 according to dump_format()
[13:15] <divVerent> it is very high quality for online radio standards
[13:15] <tuxx_>   Duration: 00:00:00.00, start: 0.000000, bitrate: -2147483 kb/s
[13:15] <tuxx_> http://libvncserver.sourceforge.net/doc/html/vnc2mpg_8c-example.html
[13:15] <divVerent> also, I am pretty sure it beats FM radio
[13:15] <tuxx_> why would i have a negative bitrate? :(
[13:15] <divVerent> tuxx_: that probably mans unknown
[13:15] <Yulth> divVerent: AAC at 64kbps? beats FM?????
[13:16] <tuxx_> divVerent: mans unknown?
[13:16] <divVerent> Yulth: FM is crap :P
[13:16] <divVerent> Yulth: or rather
[13:16] <divVerent> too many stations, too little bandwidth
[13:16] <divVerent> hehe, actually, this is an interesting question... what's the signal bandwidth according to Shannon for a FM radio station
[13:17] <Yulth> I don't know
[13:17] <divVerent> well, I am in Germany, and FM radio transmission often makes trouble
[13:17] <divVerent> when driving for example
[13:19] <Yulth> naturally
[13:19] <divVerent> okay, if I see this right on wikipedia, a single audio channel for FM radio has a bandwidth of 15kHz
[13:20] <tuxx_> divVerent: what did you mean by "that probably mans unknown"
[13:20] <divVerent> now for "perfect quality", we would want to transmit 48000/16/2 audio, i.e. 1536kbit/s
[13:20] <divVerent> tuxx_: it doesn't know the length
[13:20] <divVerent> why, I don't know
[13:21] <tuxx_> divVerent: hmm how is the duration assessed? by the number of frames / framerate?
[13:21] <divVerent> tuxx_: no idea
[13:21] <darkstarbyte_> I have encoded dvd quality video at lower bit rates
[13:21] <divVerent> normally this info comes from the container
[13:21] <tuxx_> mpg doesnt have a container tho, does it?
[13:21] <tuxx_> i mean.. mpeg does need a container
[13:21] <darkstarbyte_> mpg and mpeg
[13:22] <divVerent> just try another container
[13:22] <tuxx_> if i use avi then it knows the duration but it plays back too fast
[13:22] <tuxx_> a 10sec recording plays back in roughly 7 seconds
[13:23] <tuxx_> thats with mplayer.. vlc actually plays back the entire video in less than half a second
[13:25] <divVerent> anyway, just calculated
[13:25] <divVerent> for FM radio to be lossless, you need a SNR of 22.23dB... that should actually be achievable
[13:26] <divVerent> then the real issue is not the information rate, but the non error resilient analog coding :)
[13:29] <divVerent> tuxx_: try a container like mkv or nut
[13:29] <Yulth> I understand, but, I bet that the online radio stations don't care this concept
[13:30] <divVerent> Yulth: the thing is, online radio stations want to stay compatible with most internet connections
[13:30] <divVerent> and also work in background and not interfere with anything else you may do
[13:30] <divVerent> thus they keep their bandwidth quite low
[13:30] <divVerent> it's not meant as hifi audio transmission
[13:32] <Yulth> In your opinion, at what bitrate must be converted any raw audio file (like wav, for example) to achieve HE-AACv2 with CD quality?
[13:33] <Kuukunen> Yulth: for one I think HE-AAC implies SBR always, which means it can't do CD quality
[13:33] <Kuukunen> correct me if wrong
[13:37] <tuxx_> mkv gives me something ridiculous   Duration: 00:00:00.22, start: 0.000000, bitrate: N/A
[13:38] <durandal_1707> tuxx_: how you created mkv
[13:38] <tuxx_> durandal_1707: i am using this example code
[13:38] <tuxx_> http://libvncserver.sourceforge.net/doc/html/vnc2mpg_8c-example.html
[13:38] <tuxx_> durandal_1707: its an example of libvncserver which records a vnc session to an mpg using libavcodec/libavformat
[13:38] <tuxx_> in theory at least :)
[13:39] <Yulth> Kuukunen: Kuukunen
[13:39] <Yulth> Kuukunen: Most companies that develop codecs claim that HE-AAC at 48kbps is equivalent to CD quality...., I assume they are lying but is impactant that they still doing that...
[13:40] <divVerent> Yulth: haha, fun claim
[13:40] <tuxx_> maybe i am losing frames (i.e. not getting 25 frames per seconds) from the vnc server
[13:47] <Kuukunen> Yulth: well, "CD quality" is not very well defined :P
[13:48] <Kuukunen> Yulth: if you're happy with how HE-AAC at 48 kbps sounds to you, be my guest, use it
[13:48] <Kuukunen> it all depends on the people, the equipment, the encoder, etc etc
[13:48] <Kuukunen> and of course the source too
[13:53] <Yulth> I understand
[13:54] <divVerent> I just tested with libfdk_aac
[13:54] <divVerent> it's roughly ok, but minor issues
[13:54] <divVerent> somewhat like mp3 128k
[13:54] <Yulth> ok, and Is there any difference between libfdk_aac and other commercial codecs?
[13:54] <divVerent> it seems better than neroAacEnc
[13:55] <Yulth> O_O
[13:55] <divVerent> but tried no other one
[13:55] <divVerent> ffmpeg -i 01.wav -codec libfdk_aac -profile:a aac_he_v2 -b:a 48k out.aac
[13:55] <divVerent> this is how I tested
[13:55] <divVerent> it's a bit tricky to force ffmpeg+libfdk_aac to actually use AAC HE v2
[13:56] <divVerent> but here you have my working command line
[13:56] <Yulth> yes is the same command I use
[13:57] <divVerent> compare: http://rm.sudo.rm-f.org/img/uploaded/313b29e1ec200b6eb482dbd30dd4d741.png http://rm.sudo.rm-f.org/img/uploaded/e58a1a4c4ec3e33257923b388ae357b2.png
[13:57] <divVerent> note quite much artifacting in the upper frequency regions above 12k
[13:57] <divVerent> this is AAC HE v2
[13:57] <divVerent> I really wouldn't trust a codec doing that :P
[13:58] <divVerent> the 20.5kHz cutoff on the other hand is great
[13:58] <divVerent> what worries me most in the artifacting above is that it doesn't even depend on the pitch of what is played
[13:58] <divVerent> so I am VERY sure you can construct audio input for which this is TOTALLY obvious breakage
[13:59] <durandal_1707> divVerent: how you created that image?
[13:59] <divVerent> sox :P
[13:59] <durandal_1707> cmdline
[13:59] <divVerent> mp3graphs.sh *.aac
[13:59] <divVerent> ;)
[13:59] <divVerent> https://gist.github.com/4017081
[14:00] <divVerent> so, to construct a really bad input for this codec and implementation, I'd try a sawtooth wave at about 330Hz ;)
[14:00] <divVerent> so that the overtones hit the dark spots in the spectrum above
[14:00] <divVerent> and compare to a like 340Hz one
[14:00] <divVerent> where these would not hit
[14:01] <divVerent> I'd bet these would have quite different timbres
[14:02] <divVerent> actually, I meant 660Hz and 680Hz
[14:02] <divVerent> also, the original input had a lot of stuff at 10kHz, which the encoded audio lost entirely
[14:02] <divVerent> no idea what it is and whether one can hear it, though ;)
[14:08] <durandal_1707> you cant hear with eyes
[14:09] <divVerent> of course
[14:09] <divVerent> I can ABX these two, though
[14:10] <divVerent> the spectrum mainly tells us how much information is lost, how important this information is it doesn't say
[14:10] <divVerent> it also tells me VERY clearly that the output of the enocder is totally unsuitable for further editing after encoding
[14:11] <divVerent> or transcoding ;)
[14:11] <divVerent> because no other codec has similar spectral loss characteristics
[14:12] <divVerent> http://rm.sudo.rm-f.org/img/uploaded/edd8aae6dedd342cc78b1dcd0841e3cd.png is same input, but output is MP3 128kbps instead
[14:12] <divVerent> http://rm.sudo.rm-f.org/img/uploaded/33e7a2c4cfbf4ade24386983f7188599.png is this thing AAC-HEv2 encoded
[14:13] <divVerent> quite obviously, it's a bad idea to AAC MP3'd input :P
[14:14] <durandal_1707> i want to generate those images with ffmpeg
[14:15] <mads-> I tried using ffmpeg to record my screen over the weekend. It works fine for smaller intervals like 10 seconds - 3 minutes. It crashed after 6 minutes on my computer, looking like it was completely memory starved. This is the script I'm running: http://ideone.com/AqjCh1
[14:16] <mads-> Can anyone give me some criticism as to what I'm doing wrong
[14:40] <cbsrobot> damit
[14:40] <cbsrobot> why is it so hard to make a goodlooking webm ?
[14:40] <cbsrobot> I always get artefacts
[14:40] <cbsrobot> anybody has a good advice ?
[14:41] <ubitux> quality factor settings for libvpx seem to only slow down the encode, for no better quality
[14:41] <ubitux> maybe it needs two pass, no idea
[14:41] <ubitux> i'm only able to get good quality by setting a high bitrate
[14:42] <cbsrobot> I set it to 8m and still get artefacts
[15:09] <Mavrik> cbsrobot: I've found that setting qmax/qmin helps
[15:09] <Mavrik> cbsrobot: https://www.virag.si/2012/01/webm-web-video-encoding-tutorial-with-ffmpeg-0-9/
[15:09] <Mavrik> this gave me good results
[15:11] <cbsrobot> Mavrik: thanks
[15:12] <cbsrobot> I had -qmin 0 -qmax 10 and still al ot of artefacts
[15:12] <cbsrobot> well it was a synthetic image (graphic) not a real film
[15:12] <Mavrik> usual range for those is 0-60 for h.264
[15:12] <Mavrik> so that probably did destroy your image quality
[15:13] <Mavrik> don't corner your encoder by too strict settings :)
[15:13] <cbsrobot> I selved it with the head through the wall: gradfun
[15:13] <cbsrobot> *solved
[15:13] <Mavrik> also make sure your libvpx is uptodate
[15:14] <cbsrobot> hmmm - didn't think about that
[15:14] <cbsrobot> I'll try it
[15:15] <cbsrobot> and the qmin/qmax limitation came only after a lot of other tweaks I tried &
[15:20] <cbsrobot> libvpx v1.1.0 should be recent enough
[15:37] <tuxx_> hey guys...
[15:38] <tuxx_> can someone tell me how to drop frames?
[15:38] <tuxx_> using libav*
[15:39] <Mavrik> drop?
[15:49] <zap0> yeah boi!  drop the frames!!
[15:54] <tuxx_> Mavrik: yea im encoding a video in real time but my cpu isnt always fast enough so i would like frames to be dropped
[15:55] <Mavrik> tuxx_: basically... you just don't pass the frame to encoder
[15:55] <Mavrik> that's it :D
[15:55] <tuxx_> Mavrik: but then i lose frames
[15:55] <Mavrik> doh.
[15:55] <tuxx_> Mavrik: and the video plays faster than it should
[15:55] <Mavrik> that's the point
[15:55] <tuxx_> because it wants to play at a fixed frame rate
[15:56] <Mavrik> then there's something wrong with your frame timestamp calculation
[15:56] <tuxx_> Mavrik: i think, i dont have a frame timestamp calculation :)
[15:56] <Mavrik> which format are you encoding to?
[15:57] <tuxx_> Mavrik: hmm mpeg1/2
[15:57] <tuxx_> Mavrik: http://libvncserver.sourceforge.net/doc/html/vnc2mpg_8c-example.html
[15:57] <Mavrik> uum
[15:58] <tuxx_> Mavrik: that example C program encodes video recorded from VNC into mpeg on the fly
[15:58] <tuxx_> however when i played it back it played way too fast
[15:58] <Mavrik> hmm, MPEG-2 needs you to have a constant framerate
[15:58] <Mavrik> so you'll have to check ffmpeg.c how framedrop is done there
[15:59] <tuxx_> then i noticed it was because i was not getting 25fps so i modified it with a thread that ensures that i DO write off 25 fps
[15:59] <tuxx_> now it plays at normal speed
[15:59] <tuxx_> but this program is supposed to run on an embedded device which doesnt have a whole lot of cpu power
[15:59] <tuxx_> so i will probably need to drop some frames occassionally
[15:59] <tuxx_> when i try reducing the frame rate of mpeg from 25 to 16 i get: [mpeg2video @ 0x659b10]MPEG1/2 does not support 16/1 fps
[16:00] <tuxx_> no idea what thats all about
[16:00] <Mavrik> yeah
[16:00] <Mavrik> MPEG-2 supports just some fixed rates
[16:00] <Mavrik> that's why most people use H.264 now :)
[16:00] <tuxx_> Mavrik: is it very cpu intensive?
[16:01] <tuxx_> or rather.. which codec would you recommend for a cpu weak device.. i dont care much about quality
[16:01] <Mavrik> hmm
[16:01] <jthomas_> I've tried to RTFM and searched Google but I can't figure out how to tunnel an RTP stream over HTTP (or preferrably HTTPS).  Can anyone work with me a bit on this?  I'm running ffmpeg 1.0 on Debian Sid, trying to stream my soundcard.  I have rtp multicast working but our network doesn't allow that, so i want http
[16:01] <jthomas_> this is my RTP command, which works on a less restrictive network:   /usr/bin/ffmpeg -f alsa -i pulse -acodec libmp3lame -ab 32k -ac 1 -re -f rtp rtp://234.5.5.5:1234
[16:01] <Mavrik> I would recommend using a hardware H.264 encoder
[16:01] <Mavrik> jthomas_: I'm quite sure ffmpeg doesn't support that
[16:01] <Mavrik> you'll have to write your own proxy
[16:01] <tuxx_> Mavrik: hmm are you familiar with the beagleboard?
[16:02] <Mavrik> not really
[16:03] <tuxx_> okay.. its a rather common embedded linux device.. it has a cortex-a8...
[16:03] <Mavrik> but using an ARM processor without hardware video encoder will result in VERY poor results
[16:03] <Mavrik> those just aren't fast enough for anything decent
[16:03] <jthomas_> Mavrik on various webpages and in the ffmpeg manpage, it's mentioned but no examples are given
[16:03] <tuxx_> Mavrik: ugh.
[16:03] <tuxx_> Mavrik: alright thanks for your help mate...
[16:04] <Mavrik> tuxx_: you can try doing libx264 encode with -preset ultrafast
[16:04] <Mavrik> and see what you get out
[16:04] <Mavrik> and make sure you compile with NEON support
[16:09] <mads-> How do I pipe all the output from ffmpeg to a log file? doing ffmpeg -i input .... output.avi 2>&1 > ffmpeg.log doesn't seem to do the trick
[16:15] <DonGnom> mads-: ffmpeg -i input .... output.avi > ffmpeg.log 2>&1
[16:16] <mads-> DonGnom, thanks. What is the difference between those?
[16:18] <DonGnom> mads-: thats how it workes for me (think you also could use ffmpeg -i input ... output.avi &> ffmpeg.log
[16:18] <mads-> DonGnom, it works for me as well. I was just curious if you knew the difference between the two :)
[16:18] <DonGnom> the ordering :)
[16:19] <DonGnom> bash is really mad when something doesnt come in that order it wants :D
[16:19] <mads-> I just thought telling it to pipe stderr to stdout before where to pipe stdout was the logical order
[16:22] <DonGnom> mads-: yeah little bit unlogical but it works *shrug*
[17:21] <mudkipz> While looking around I heard about something called "gdigrab" which is/was apparently supposed to be an implementation for a windows version of 'x11grab' like 'vfw' and 'dshow'. Did this ever get implemented?
[17:22] <mudkipz> Also, dshow is still the best way to capture now, right?
[17:26] <knoch> ubitux: is it possible to make ffmpeg quit if a live stream input is bad ? thank you
[17:27] <msmithng> how do you define 'bad'
[17:28] <msmithng> knoch ^
[17:28] <ubitux> why me? :(
[17:28] <ubitux> look at -err_detect
[18:03] <mudkipz> I'm pipelining one ffmpeg command into another command and getting a negative quantizer in the output of one of them.
[18:03] <mudkipz> Like this
[18:03] <mudkipz> frame=105998 fps= 25 q=13.0 size= 261770kB time=01:11:22.40 bitrate= 500.8kbits
[18:03] <mudkipz> frame=105998 fps= 25 q=-1.0 size= 261813kB time=01:11:22.38 bitrate= 500.8kbits
[18:03] <mudkipz> What does a negative q value mean?
[18:10] <bizulk> Hi ! I played an h264 4840p video on ARM omap platform (Beaglboard) using omapfbplay. As a get twice the CPU load of the gstreamer-ti plugin I'm not sure that DSP acceleration is supported by the libavcodec. And by reading some mailing list I really doubt it. Can somebody confirm this plz ?
[18:13] <msmithng> honestly mudkipz, in general... I've yet to see anything illustrating what the q value really means overall.
[18:13] <mudkipz> yea, I had no idea either until a little bit ago when I looked here.
[18:13] <mudkipz> http://ffmpeg.org/ffmpeg.html#Tips
[18:13] <mudkipz> Apparently it's the quantizer.
[18:14] <mudkipz> Though, I don't really understand what that means except that it's some sort of measure of the quality that could be achieved or effort exherted or something.
[18:15] <msmithng> awesome... that's new detail
[18:15] Action: msmithng tips hat
[18:16] <mudkipz> hehe, you're welcome
[18:16] <mudkipz> I still have no idea why I'd get a negative quantizer value though.
[18:16] <mudkipz> I can't tell if that's a bad thing or a good thing.
[18:46] <tuxx_> wyhsp
[19:47] <sean1988> Okay i'm back from yesterday and i think somehow the problem is ubuntu or debian.
[20:14] <kanyl> Can i specify my yasm path while compiling ffmpeg? It can't be found right now.
[20:15] <JEEB> export PATH=/herp/derp/bin:${PATH} ?
[20:16] <JEEB> and then run configure/make
[20:16] <JEEB> the PATH would be specific to that shell session
[20:20] <sean1988> any ideas why whenever i try to converts somthing i keep getting
[20:20] <sean1988> [dvd @ 0x18d1c80] buffer underflow i=0 bufi=59774 size=61446
[20:20] <sean1988> the last two number change.
[20:20] <sean1988> get it in debian and windows.
[20:21] <sean1988> also why is ffmpeg much slower on ubuntu and debian then windows and magiea.
[20:24] <JEEB> means that you went over the bitrate limits
[20:25] <JEEB> that you have set
[20:27] <sean1988> fair enough lol any ideas about the later question.
[20:30] <sean1988> I don't really want to use magiea the packet manager seems confusing, but seems the be the only place ffmpeg gets good fps
[20:33] <llogan> sean1988: hard to say without knowing the details
[20:33] <llogan> did you compile ffmpeg? are you using the same input source and options
[20:33] <llogan> are you even using the same encoder(s)?
[20:34] <sean1988> don't know about the last part somebody in here was helping me ysterday but he couldn't help
[20:34] <sean1988> I've tried from the repos and compiling myself.
[20:35] <sean1988> and yes i used the same source input and options
[20:37] <llogan> what ubuntu version? how do you know windows is faster?
[20:37] <sean1988> I'm on debian testing now but ubuntu yesterday was 12.10
[20:38] <llogan> "ffmpeg" from the repo of 12.10 is not from FFmpeg
[20:38] <sean1988> I also complied it from ffmpeg website and had same issue
[20:39] <llogan> what is the issue, exactly? that it is slower than you'd expect?
[20:40] <sean1988> yes, much slower
[20:40] <llogan> you should provide all of the necessary information so others can duplicate the issue.
[20:40] <llogan> including: your ffmpeg versions, your ffmpeg command, the complete console output, and your input file
[20:42] <meekohi> I'm having trouble using av_open_input_file inside a templated class: http://pastebin.com/d38A9U8E
[20:42] <meekohi> Anyone understand what's going on here?
[20:43] <llogan> meekohi: you might have better luck at the libav-user mailing list: http://ffmpeg.org/mailman/listinfo/libav-user
[20:43] <meekohi> llogan: Ah okay thanks.
[20:43] <llogan> not many API answers here, but you might get an answer if you wait
[20:45] <meekohi> Oh I'm always up for waiting.
[20:57] <sean1988> okay where do I post all this information?
[20:59] <ubitux> pastie, pastebin, etc.
[21:03] <bruce> is it possible to create the moov atom of an h.264 encoded .mp4 file if encoding was killed before the moov atom was created?
[21:03] <sean1988> should i wait for it to compete or can i cancel it?
[21:07] <Leoneof> hello, just curious question, why corrupted mp4 file doesn't work? isn't the information about resolution, bitrate...etc are in the header of mp4 file?
[21:12] <MP4_maniac> Leoneof: for some codecs such as mpeg-4 video/audio, the decoder specific info which must requires decoding is stored in moov atom. so, if the moov atom in a mp4 file is corrupted, that mp4 file might not work
[21:14] <JEEBsv> moov atom basically contains the overall structure of the file, so if it's broken you can't really do anything about it
[21:14] <JEEBsv> (unless you use the fragments feature, but that's not really used by pretty much anyone)
[21:14] <Leoneof> MP4_maniac: ah indeed, i've seen the error about moov atom
[21:15] <Leoneof> i see
[21:15] <JEEBsv> fragments = lots of little 'indexes' all over the file
[21:16] <MP4_maniac> JEEBsv: even if fragment, decoder specific info is always absent in each fragment.
[21:16] <JEEBsv> ah
[21:16] <MP4_maniac> stored in the initial moov atom only
[21:31] <Leoneof> thanks :)
[21:34] <sean1988> http://pastie.org/5191673
[21:39] <cbsrobot> sean1988: not sure what you do, but keep it simple
[21:39] <cbsrobot> ffmpeg -i input.avi -qscale 0 -target pal-dvd output.mpg
[21:39] <cbsrobot> or -qscale 1
[21:40] <sean1988> still getting low fps
[21:41] <cbsrobot> how low is low ?
[21:42] <sean1988> 170 but on windows i get 320 and madriea i push around 450
[21:44] <bruce> so, without a moov atom, a file can't be decoded even if you know the original settings used to encode the file? (and the video stream being the only stream in the file... ie. no audio to interleave)
[21:45] <cbsrobot> sean1988: can you pastebin ffmpeg loglevel debug -i input.avi -qscale 0 -target pal-dvd output.mpg
[21:45] <cbsrobot> ah sorry
[21:45] <cbsrobot> ffmpeg -loglevel debug -i input.avi -qscale 0 -target pal-dvd output.mpg
[21:45] <JEEBsv> bruce: it would still be 'fun' to try and recover
[21:46] <relaxed> 1 is the lowest
[21:46] <sean1988> okay but do i have to convert the whole video or can i  stop it?
[21:46] <JEEBsv> you can seemingly try to mimic and check if it matches to something that a certain muxer does
[21:46] <JEEBsv> like libavformat, GPAC or L-SMASH
[21:46] <cbsrobot> stop it
[21:47] <llogan> L-SMASH is in arch AUR if by chance you're using that distro
[21:47] <sean1988> http://pastie.org/5191812
[21:47] <bruce> well, I happen to have another file that was encoded with exactly the same settings at exactly the same time, but from another source.... perhaps a moov atom transplant?
[21:48] <bruce> the files are different though... the one that died did so a few minutes early, and is much larger as the source video signal has more noise.
[21:49] <JEEBsv> you could see if there are similarities in muxing :s
[21:49] <JEEBsv> but you'd probably have to learn about the internals
[21:49] <JEEBsv> of the format
[21:51] Action: llogan would re-encode
[21:51] <cbsrobot> sean1988: detected 8 logical cores ?
[21:51] <sean1988> amd bulldozer
[21:53] <cbsrobot> on exactly the same hardware ?
[21:53] <cbsrobot> same disks ?
[21:53] <sean1988> yup
[21:53] <llogan> where did you get your windows ffmpeg binary?
[21:54] <cbsrobot> and this is & ubuntu ?
[21:54] <sean1988> ffmpeg website
[21:54] <sean1988> debian
[21:54] <sean1988> but had same issue on ubunty
[21:54] <sean1988> ubuntu*
[21:54] <llogan> ffmpeg website does not distribute ffmpeg binaries
[21:54] <sean1988> i didn't compile it
[21:54] <llogan> what you say!!
[21:54] <sean1988> http://ffmpeg.zeranoe.com/builds/
[21:55] <relaxed> we get signal
[21:56] <llogan> for great justice
[21:58] <llogan> sean1988: you can benchmark decoding first: ffmpeg -benchmark -i input -f null -
[21:59] <sean1988> -i input?
[21:59] <llogan> might have to use out.null instead of - in Windows, but what do i know
[21:59] <cbsrobot> -i input.avi
[21:59] <llogan> "input" is the name of your input file
[21:59] <JEEBsv> /dev/null on windows is just NUL
[21:59] <sean1988> input.aci
[21:59] <sean1988> avi*
[22:00] <sean1988> I'm in debian atm.
[22:01] <llogan> either way, so you can compare on both
[22:02] <sean1988> http://pastie.org/5191892
[22:03] <llogan> ffmpeg -benchmark -i input.avi -f null -
[22:04] <llogan> then in windows: "ffmpeg -benchmark -i input.avi -f null NUL" or "ffmpeg -benchmark -i input.avi -f null foo.null"
[22:04] <llogan> compare times
[22:05] <sean1988> im getiing 1500 fps?
[22:05] <llogan> if the times are the same then you know the issue (probably) hasn't to do with decoding
[22:06] <sean1988> http://pastie.org/5191912
[22:07] <llogan> or i should say "if the times are similar"
[22:08] <sean1988> okay i'm going to load up my windows partion i'll be back with a paste bin of ffmpeg.
[22:16] <llogan> that's a long boot
[22:17] <sean1988> http://pastie.org/5191941
[22:17] <relaxed> main screen turn on
[22:18] <sean1988> nearly hit 2500 just before half way
[22:20] <llogan> now go back to linux.
[22:20] <llogan> just kidding
[22:20] <sean1988> i am in linux
[22:20] <sean1988> i don't have a windows irc client downloaded
[22:22] <llogan> hmm...maybe that wasn't such a useful test.
[22:23] <llogan> but it does show that you are decoding faster in windows.
[22:24] <sean1988> so people won't think i'm crazy
[22:25] <sean1988> im tempted to install mageia again to do the test there just to see the results.
[22:26] <sean1988> but it took my forever to install last night
[22:26] <cbsrobot> sean1988: do you want to build ffmpeg on ubuntu yourself and see if it makes a difference ?
[22:26] <sean1988> I have done, I'm on debian atm
[22:32] <llogan> i'll see if there is a difference on this machine next time i'm in windows, although last time i tried it a few years ago windows was slightly slower
[22:32] <sean1988> what os are you using?
[22:32] <llogan> arch linux/win 7
[22:32] <sean1988> i was going to try and arch install but my disc just wouldn't boot =/
[22:33] <llogan> welcome to arch
[22:33] <sean1988> Lol.
[22:33] <sean1988> booted on my old machine.
[22:34] <llogan> actually i've encountered two issues on my old machine where it freezes completely due to 1) something with the kernel, and 2) an openchrome issue
[22:36] <sean1988> any ideas on how i would get the arch disc to actually load the os? i couldn't figure out how to get onto the arch irc, i'm new to irc.
[22:36] <llogan> /join #archlinux
[22:37] <llogan> might have to be a registered IRC nick though. i can't remember
[22:37] <sean1988> yeah you do and that's what i don't know how to do.
[22:37] <llogan> http://freenode.net/faq.shtml#contents-userregistration
[22:39] <llogan> let's just blame debian/ubuntu and call it a day. good luck with that.
[22:40] <sean1988> lol yeah i think debian is at fault, and the problem translates over to ubuntu
[23:32] <stephanedev> hi everyone, i have a question regarding writing a packet to different streams
[23:32] <stephanedev> according to the doc for av_interleaved_write_frame, libavformat takes ownership of the packet data and thus it should not be accessed after the function returns
[23:33] <stephanedev> my initial idea was to loop on my streams and write the packet with successive calls to av_interleaved_write_frame
[23:33] <stephanedev> but i guess i cannot do that given the function description
[23:34] <stephanedev> should i have one copy of the packet per stream?
[23:35] <Mavrik> stephanedev, yeah, otherwise you'll have a hell to pay :)
[23:35] <Mavrik> stephanedev, because when first of the muxers frees the packet the second will crash :)
[23:36] <stephanedev> Mavrik: thank you for your help once again. yes that's what i thought after reading the doc
[23:37] <stephanedev> is there any recommended way to copy a packet? I tried to look for a function to do that but looking for *copy* didn't bring any interesting result
[23:37] <burek> http://www.hongkiat.com/blog/cinemagraph/
[23:39] <Mavrik> stephanedev, av_dup_packet?
[23:44] <stephanedev> Mavrik: thanks. I searched again after reading the doc for av_dup_packet (the doc says it's a hack) and actually there is a function called av_copy_packet
[23:45] <stephanedev> but i am not too sure about the destruction behavior when looking at its implementation
[23:46] <stephanedev> the destination packet destruct field is set to av_destruct_packet, i would have expected to just copy src->destruct instead
[23:57] <Pupuser-2> hello: http://pastebin.com/7HuiLKW9
[23:57] <Pupuser-2> why this code does not works??
[23:57] <Pupuser-2> does not convert files from mp4 to mp3
[00:00] --- Tue Nov  6 2012