[Ffmpeg-devel-irc] ffmpeg.log.20140625

[burek]

[00:00] <_aeris_> for 4h video, not able to run this command :P
[00:04] <debianuser> Maybe something like extracting audio, joining audio, concatting video, and muxing them together would work for you...
[00:05] <debianuser> Or maybe (since your second audio is actually silence) you can join videos, and insert the audio with nonzero start pts so that it started where you need it to...
[00:06] <debianuser> anyway I have no idea how to do that. :)
[00:11] Action: debianuser looks at `-itsoffset offset` option.
[00:15] Action: debianuser now looks at https://trac.ffmpeg.org/ticket/1349 - The -itsoffset option does not work the way you think it does.
[00:39] <debianuser> _aeris_: This had just worked for me too: ffmpeg -filter_complex aevalsrc=0:d=3 -ar 24000 -ac 1 0.ogg; ffmpeg -i "concat:0.ogg|mono24kHz.ogg" -c copy cat.ogg; mplayer cat.ogg
[01:03] <Technicus> Hello, I am looking for suggestions of Linux compatible video capture cards that have HDMI input, and connect to USB 3.0?
[01:04] <Hello71> 1. i don't think we do hardware suggestions
[01:04] <Hello71> 2. pick anything that's uvc and has hdmi input
[01:04] <Technicus> Hello71: I am just asking.
[01:06] <Technicus> Given the nature of FFmpeg I suspect that someone may have experience with such devices.  Maybe I am totaly wrong . . .
[01:08] <Hello71> just pick anything that supports uvc
[01:08] <Hello71> which should be almost all usb devices
[01:23] <Technicus> Hello71: Thanks.
[03:36] <debianuser> What are the ranges for XX in -aq XX/-q:a XX/-qscale:a XX? Also do I understand it correctly, that all three options are just aliases?
[03:41] <relaxed> debianuser: yes, and the ranges depend on the encoder
[03:41] <debianuser> Is there a way to "ask" decoder for its ranges?
[03:42] <relaxed> you mean encoder
[03:42] <relaxed> which one are you using?
[03:42] <debianuser> libvorbis right now
[03:43] <debianuser> But I'm also curious how to find it for other encoders to
[03:43] <relaxed> man oggenc | less +/'  -q'
[03:44] <relaxed> look at man lame for its ranges
[03:44] <debianuser> what about e.g. fdkaac?
[03:45] <debianuser> or for ffmpeg's own vorbis/aac encoders?
[03:48] <relaxed> libfdk_aac uses -vbr, see "ffmpeg -h encoder=libfdk_aac"
[03:49] <relaxed> do not use ffmpeg's vorbis encoder
[03:49] <relaxed> I don't think the native aac encoder suports vbr
[03:50] <relaxed> you should be using libvorbis or libfdk_aac
[03:51] <debianuser> I think I've got the idea. Thank you!
[04:04] <relaxed> you're welcome
[04:13] <Hello71> info for popular codecs is also on trac wiki
[04:13] <Hello71> e.g. https://trac.ffmpeg.org/wiki/vpxEncodingGuide
[04:25] <wookienz> hi guys, just trying to use ffmpeg to re encode to aac and h264. The resulting file appears to be a lot lower bit rate. What flags can i use to ensure that the but rate is as high as possible? thanks.
[04:36] <debianuser> wookienz: You mean the file has lower bitrate (i.e. it's smaller) but still a good quality?
[04:38] <wookienz> debianuser, medininfo shows a lower bitrate, i just assume it is a lower quality. im a noob when it comes to this!
[04:42] <debianuser> Modern codecs are very complex and had lots of settings. The same codec with the same bitrate may have much better quality depending on its other settings. And the other way around: same quality may be achieved on different bitrates depending on used encoding profiles and other options.
[04:43] <wookienz> hmm ok. back to square one.
[04:45] <debianuser> A simple encoding command: ffmpeg -i in.avi -c:v libx264 -c:a libfdk_aac out.mp4
[04:45] <debianuser> can be tuned with lots of different options: ffmpeg -i in.avi -c:v libx264 -crf 22 -r 25 -preset veryslow -tune grain -profile:v baseline -level 3.0 -pix_fmt yuv420p -movflags faststart -c:a libfdk_aac -b:a 128k -ac 2 -cutoff 18000 out.mp4
[04:45] <debianuser> These links may be interesting for you: https://trac.ffmpeg.org/wiki/Encode/H.264 https://trac.ffmpeg.org/wiki/Encode/AAC
[04:50] <debianuser> But to make the long story short: usually it's not "bitrate", but "quality" is controlled. In the above example video quality is controlled by the option -crf 22 with range (lossless)0..51(worst), default is 23. From that option codec automatically calculates and adjusts the bitrate to get the requested quality.
[04:51] <debianuser> Another thing that affects bitrate/quality a lot is "profile" (baseline, main, high...) Higher profile gives better quality for the same bitrate. But you have to consider compatibility: many iphone/android players can only handle "baseline" profile.
[04:51] <wookienz> ok thanks, ill have a look into those.
[04:52] <wookienz> ok, sounds like they are good easy presets
[04:53] <debianuser> As for what quality to target... Try different options and look at them side-by-side, if the quality is too good  you can lower the "-crf", if it's too bad  make it higher. I don't know any better way. :(
[04:54] <wookienz> unfortunately i  am looking at converted hundreds of files so wont be able to side by side them.
[04:57] <debianuser> Maybe experiment with a few, to get the idea of how much the quality is changed, then add some safety margin, and use those settings for the others. :) If you don't have time for experiments, use default values, they _should_ be safe.
[04:58] <wookienz> default as in: ffmpeg -i file.mkv -c:a libfdk_aac -c:v copy out.mp4 ?
[05:13] <debianuser> If the input is already H.264 and compatible with whatever you want it to be compatible then yes. By the way, libfdk_aac by default cuts frequencies higher than 14kHz (it's not ffmpeg's default, it's libfdk_aac's one). So if you're encoding some high quality audio you may want to add option "-cutoff 20000" after "-c:a libfdk_aac".
[05:18] <wookienz> ok, thanks.
[05:24] <thshdw> I know that ffmpeg is able to encode and decode in atrac3, and it can also read some metadata like artist, year of track etc. Can it also write this metadata to atrac files?
[11:44] <debianuser> Heh. Video encoder for 8088 CPU: http://trixter.oldskool.org/2014/06/19/8088-domination-post-mortem-part-1/
[11:47] <ubitux> there is no lowres support with h264?
[11:48] <iive> thank god, no
[11:48] <ubitux> too bad :)
[11:48] <iive> not really
[11:50] <iive> last time i did test with lowres, the 2^3 reduction of resolution resulted in astonishing 30% speedup.
[11:50] <ubitux> not that bad then
[11:50] <iive> and this also causes quite a lot of prediction errors.
[11:50] <ubitux> i don't need the output frame
[11:51] <ubitux> i'm looking at extracting just the mv somehow
[11:51] <ubitux> like avframe side data or some other mean
[11:51] <iive> well, grayscale decoding gives you same speed up and full resolution and no artifacts.
[11:51] <ubitux> but then i wouldn't want the frame buffer anyway, just the side data
[11:52] <ubitux> maybe i should just parse the avpacket myself, dunno
[11:52] <iive> there are flags for skipping decoding, they might be more useful...
[11:52] <iive> and there should also be some parsers
[11:52] <ubitux> ah?
[11:53] <ubitux> there are some decode skipping flags?
[11:53] <iive> how else would you do a frame drop?
[12:23] <cuba> can anyone tell me why my resulting file has no video length in vlc?
[12:23] <cuba> http://codepad.org/42vVeFCJ these are the arguments to create it
[12:26] <c_14> What value does (fps) have? And, does ffprobe/ffplay show the correct video length?
[12:30] <cuba> c_14: fps is the frames per seconds of the input coming in
[12:30] <cuba> Duration: N/A :(
[12:38] <tfirg> i would like some suggestions to improve the quality of my x11 grab. I use this command: "ffmpeg -y -f x11grab -draw_mouse 1 -framerate 5 -video_size 3840x2160 -i :0.0 -f alsa -i pulse -c:v libx264 -pix_fmt yuv420p -b:v 2500k -c:a aac -strict -2 -b:a 192k -preset ultrafast -tune zerolatency -crf 18 -movflags +faststart testgrab.mp4"
[12:39] <tfirg> the result is here (2 minute video) : https://www.youtube.com/watch?v=dt-WiII6Z2A
[12:39] <tfirg> it shows blue block and flashes
[12:39] <tfirg> and some other glitches
[12:40] <tfirg> is there a way to make that better?
[13:16] <c_14> cuba: I got that, what I meant is that you're setting it to a value somewhere, what value are you setting it to.
[13:18] <cuba> 20 atm
[13:18] <cuba> i think its mayabe that my ffmpeg process isnt directly terminated after recording stops
[13:18] <cuba> i terminate in destructor of the object
[13:19] <cuba> but the options that way are correct and shoudl result in a proper duration?
[13:22] <cuba> mh not really the case it seems
[13:24] <c_14> you might want to switch the input option to -framerate instead of -r, I've had some issues with that.
[13:26] <cuba> ill try that
[13:27] <cuba> thanks c_14
[13:36] <c_14> Does it work?
[13:38] <cuba> no
[14:04] <cuba> c_14: it isnt ffmpegs fault
[14:05] <cuba> thanks for your help
[14:07] <tfirg> http://fpaste.org/113009/36979891/ anyone? any little suggestion/pointer appreciated
[14:10] <c_14> Increase your framerate? Also, setting a bitrate and a crf is useless with x264 (afaik). maybe try with git ffmpeg (assuming you aren't).
[14:14] <tfirg> thanks yes i will gradually increase framerate , i suspect it might cause buffer underruns
[14:14] <tfirg> as for the bitrate and crf, i suppose it wont hurt either
[14:14] <tfirg> not using git, using 2.1.4
[14:15] <tfirg> the biggest problem is the blue blocks and flickers
[14:43] <termos> When using ffplay on a stream it will say bitrate: N/A, is this because the stream doesn't want to supply this metadata?
[14:44] <termos> I'd like to know the input bitrate to my encoder, but the av_dump_format will also tell me bitrate: N/A
[15:22] <metathink> hi
[15:55] <leeivan86> is there anyway to avoid reencoding a video when i just want to crop the video to get rid of blackboxes?
[15:55] <sacarasc> No.
[15:56] <leeivan86> ok thanks
[15:56] <relaxed> leeivan86: mplayer and friends have filters for this
[15:56] <leeivan86> ah ok
[15:57] <relaxed> in fact, most of ffmpeg's filters were ported from mplayer
[16:47] <jkli> hi guys
[16:48] <jkli> i want to batch convert videos but want also to cut / normalize to null / silence certain parts in the video
[16:48] <jkli> any idea?
[16:49] <klaxa|work> write a shell-script?
[16:50] <jkli> how do i edit the audio
[16:51] <relaxed> using sox would work
[16:51] <relaxed> I'm not sure if ffmpeg has a filter do that.
[16:52] <jkli> no, i want to silence an audio part
[16:52] <jkli> without cutting the video
[16:56] <relaxed> demux the audio, gogogadget sox, remux the audio back with the video
[16:58] <relaxed> ah, look at ffmpeg's afade filter
[17:04] <jkli> ah
[17:04] <jkli> afade is what im looking for
[17:28] <Neppy> hi there
[17:30] <Neppy> trying to use ffmpeg :D for some reason i have no avformat_seek_frame function.. so using avformat_seek_file (the docs claim thats still WIP though? so i'm confused..) however that hands me error code -1, which i can't really seem to find much about
[17:32] <klaxa|work> there is no avformat_seek_frame in the first place
[17:32] <klaxa|work> how are you calling avformat_seek_file? did you check what -1 means?
[17:33] <klaxa|work> did you try to use av_seek_frame?
[17:33] <Neppy> well tried to check what it means, didnt find any info on that and av_strerror  just says "Error number -1 occured"
[17:34] <Neppy> yeah will probably try av_seek_frame and calling it wiht: int64_t timestamp = frameNumber * AV_TIME_BASE;
[17:34] <Neppy>         int error = avformat_seek_file(avContext, videoStreamIndex, timestamp, timestamp, timestamp, AVSEEK_FLAG_FRAME);
[17:35] <Neppy> same for av_seek_frame... i think my timestamp idea may be off
[18:03] <Neppy> yeah thats the issue..
[20:51] <ChocolateArmpits> Has anyone compiled ismindex here ?
[23:02] <lanodan> I'm trying to convert mp4 to ogv but I got so much pixelisation even with -qscale 0 and -qmax 0
[00:00] --- Thu Jun 26 2014