[Ffmpeg-devel-irc] ffmpeg.log.20150806

burek burek021 at gmail.com
Fri Aug 7 02:05:01 CEST 2015

[00:06:54 CEST] <c_14> sine0: yes, why?
[01:13:16 CEST] <kyleogrg> how can i convert a "normal" color space video to 256 colors?
[01:14:30 CEST] <kyleogrg> in other words, how can i reduce the amount of colors when encoding a video?
[01:22:12 CEST] <kepstin> kyleogrg, for what, an animated gif? you'll want to look into the palettegen and paletteuse filters
[01:22:39 CEST] <kyleogrg> no, just wondering if it was possible
[01:22:54 CEST] <kepstin> for anything other than an animated gif you probably don't want to do that, it'll make the video harder to encode/look worse for most codecs.
[01:27:40 CEST] <kyleogrg> ok
[01:34:31 CEST] <durandal_1707> look at elbg filter
[01:40:50 CEST] <kyleogrg> okay
[01:42:15 CEST] <edoceo> if I've got lots of video to filter (scale, crop, layout) what can I choose as a reliable, maybe fast interim codec/format?
[01:44:24 CEST] <kepstin> edoceo, if you've got lots of disk space and io bandwidth, something like ffvhuff is ridiculously fast
[01:44:43 CEST] <kepstin> ffv1 is smaller/slower
[01:45:16 CEST] <kepstin> could also try using a lossy codec like x264 with high quality/fast encode settings
[01:45:55 CEST] <edoceo> which audio ? flac Ok?  MKV contasiner?
[01:46:18 CEST] <kepstin> sure, flac's a decent choice for audio, and any of the above codecs should work in mkv i think.
[01:48:43 CEST] <kevmitch> If you're doing processing on the audio, it's better to use an interim format that supports float
[01:49:25 CEST] <kevmitch> wav is probably easiest, but there's also wavpack
[01:50:28 CEST] <edoceo> so ffvhuff and wav in a mkv ?
[01:50:35 CEST] <kepstin> hmm, most of the time you probably won't be doing enough to audio to worry about 16bit being a limit.
[01:50:45 CEST] <kepstin> and flac can do 24bit audio which should be more than enough.
[03:48:27 CEST] <kyleogrg> hey
[03:48:45 CEST] <kyleogrg> i have a command all ready, and now i'd like to run it on a bunch of videos
[03:48:59 CEST] <kyleogrg> i've done this before a couple of different ways
[03:49:19 CEST] <kyleogrg> but is there a gui which lets you do this and shows a progress bar with estimated time remaining?
[04:46:15 CEST] <k_sze> When reading frames from a video file with ffmpeg, does it care that the reading is performed on multiple threads?
[04:48:11 CEST] <k_sze> I want to have a timer that triggers the reading of a frame, but this will be a .NET System.Timers.Timer, so the timer action will run on any available thread in a thread pool.
[04:49:10 CEST] <k_sze> Synchronizing my own timer action so that is is effectively sequential and non-reentrant is not a big problem.
[04:51:32 CEST] <k_sze> My question is more like, would ffmpeg say "Look, this file is opened in Thread 1, I can't allow you to demux and decode in Thread 2."
[09:29:22 CEST] <baadf00d> Hi, i have a question regarding distribution. Apparently I can use gpl versioned ffmpeg from my non-gpl software, if I use it through command line.
[09:29:49 CEST] <baadf00d> but can I distribute gpl ffmpeg along with my non-gpl software?
[11:51:55 CEST] <spvhere> trying to live stream webcam audio & video feed using ffmpeg and ffserver in webm format, libvorbis for audio and libvpx for video, I get a dimension not set on ffserver and conversion failed on ffmpeg
[14:37:58 CEST] <termos> when printing out the pts values converted to seconds for audio and video the video is about 2.5 seconds behind the audio. The audio and video is synced though, I'm guessing because of interleave_frame
[14:38:09 CEST] <termos> should this cause any problems?
[14:54:34 CEST] <Mavrik> only if the decoder doesn't have a small buffer
[14:54:36 CEST] <Mavrik> er.
[14:54:45 CEST] <Mavrik> only if decoder's buffer is smaller than 2.5 seconds
[14:54:51 CEST] <Mavrik> ideally the offset should be close to 0
[14:55:34 CEST] <termos> okey thanks, I don't know where this offset comes from. I'll try to see what the pts is right after the demuxing
[15:18:50 CEST] <Stifler> King of kong: fistfull of quarters ; Chasing ghosts ; TILT The battle to save pinball ;
[15:18:58 CEST] <Stifler> Best movies ever!!
[15:19:30 CEST] <Stifler> I am old :p
[15:35:07 CEST] <termos> Mavrik seems like the internal encoding buffer for video is larger than the one for audio (fdk_aac)
[20:25:23 CEST] <Nolski> Hey, could someone explain to me what the error "Error while decoding stream #1:1: Cannot allocate memory" means exactly? I see it pretty often in a variety of commands but the output normally seems fine.
[20:34:03 CEST] <llogan> the command and console output would be helpful
[20:46:43 CEST] <DHE> also if you're running other commands that would consume resources. running out of swap, having too many processes, running in 32 bit mode, etc may cause these kinds of errors. (too many processes manifests in several strange ways along the pipeline)
[20:52:09 CEST] <Nolski> DHE: llogan It's a part of a massive batch of edits on one file (that's generated by a python script).
[20:52:25 CEST] <Nolski> But here is kind of a sample of the output where it happens (check out line 53) https://gist.github.com/Nolski/40027414f49be4cc1b34
[20:54:12 CEST] <Nolski> drop keeps increasing until it hits the total number of frames and then the command ends
[21:05:29 CEST] <OstlerDev> Is there a flag that allows conversion of a file that has not finished downloading? Or on an error just try again instead of crashing?
[21:29:38 CEST] <c_14> OstlerDev: mp4?
[21:57:15 CEST] <kyleogrg> Do you think yadif is necessary in this command? http://pastebin.com/7UKiJYSK
[21:57:53 CEST] <kyleogrg> I'm converting interlaced ntsc avi video to super-low bitrate, low-res videos
[21:58:07 CEST] <kyleogrg> at half the frame rate
[21:58:42 CEST] <kyleogrg> I'm wondering if the quality is too low to notice the difference between yadif and a faster deinterlacer
[21:59:14 CEST] <DHE> well, if you're using placebo then I'd say "use the better deinterlacer" :)
[21:59:36 CEST] <kyleogrg> yeah
[22:00:06 CEST] <kyleogrg> what if i did "framestep" before "yadif"?  would that hurt anything?
[22:02:30 CEST] <edoceo> How can one fix corrupt Duration on a video?
[22:02:48 CEST] <edoceo> I've transcoded to an interim format and back but the Duration still reports 2 years
[22:03:53 CEST] <edoceo> I need an ffmpeg consultant to get in here and help with some of these issues
[22:05:00 CEST] <DHE> kyleogrg: based on the source I'd say that yadif does use multiple frame to assist in deinterlacing and you should move framestep down the filter chain
[22:05:10 CEST] <DHE> ie. leave it as is
[22:05:40 CEST] <DHE> though why are you scaling it twice?
[22:06:14 CEST] <kyleogrg> the first scaling is to convert it to square pixels (ie, 720x480 becomes 640x480)
[22:07:13 CEST] <kyleogrg> is that can be converted into one formula (i don't know exactly how) and it will make encoding faster, then i'd like to know
[22:07:24 CEST] <kyleogrg> if* that ....
[22:10:15 CEST] <klaxa> edoceo: did you try the ffmpeg-user mailing list?
[22:10:24 CEST] <edoceo> I have not yet
[22:11:00 CEST] <klaxa> i think more devs read that ML than this irc channel (because of the real-time nature of irc)
[22:11:14 CEST] <klaxa> it can't hurt to post it there too
[22:45:55 CEST] <kyleogrg> can this be combined into one scale?  scale=iw*sar:ih,scale=-2:ih/3.3333
[23:44:50 CEST] <shadowkyogre> for some reason, the ffmpeg command yields a lot sloppier sound than using arecord to record the speaker input
[23:44:51 CEST] <shadowkyogre> .asoundrc: https://ptpb.pw/G2hY
[23:44:51 CEST] <shadowkyogre> command I used to test ffmpeg audio quality with video recording: https://ptpb.pw/wAjP (I get a bunch of cannot estimate delay errors in vlc. Some poking in the #alsa channel pointed out that the yuv420p flag was causing issues. The audio is better when I remove it, but the audio still plays a bit funny)
[23:44:51 CEST] <shadowkyogre> command I used to test ffmpeg audio quality w/o video: ffmpeg -f alsa -ac 2 -ar 48000 -i dsnoop:Loopback,1  myloop.wav
[23:44:51 CEST] <shadowkyogre> command I used to test audio quality with normal arecord: arecord -fdat -D "dsnoop:Loopback,1" -d 10 test-mix.wav
[23:53:44 CEST] <OstlerDev> c_14 Yes, to mp4
[23:56:55 CEST] <c_14> OstlerDev: then no
[23:57:02 CEST] <c_14> mp4 store important metadata at the end of the file
[23:57:46 CEST] <OstlerDev> c_14 you can add the setting to put the metadata at the start
[23:57:53 CEST] <OstlerDev> Oh, you mean source file?
[23:57:58 CEST] <OstlerDev> Usually AVI or MKV
[23:58:16 CEST] <OstlerDev> I want to convert to MP4, or some other h.264 format for streaming to an Apple TV
[00:00:00 CEST] --- Fri Aug  7 2015

More information about the Ffmpeg-devel-irc mailing list