[Ffmpeg-devel-irc] ffmpeg.log.20160516

[burek]

[00:30:44 CEST] <styler2go> how can i comoile ffmpeg with stereo3d filter?
[00:31:31 CEST] <styler2go> i just compiled from git release/3.0 and get the error message: [AVFilterGraph @ 0x36b7340] No such filter: 'stereo3d'
[00:32:10 CEST] <styler2go> version: ffmpeg version n3.0.2-4-g68dcb46 and my command: ffmpeg -i Bolt\ \(2008\).mkv -vf stereo3d=sbsl:ml Bolt\2d.mkv
[00:32:59 CEST] <c_14> I think that filter is GPL so you have to --enable-gpl
[00:33:07 CEST] <durandal_170> need gpl flag to configure
[00:33:24 CEST] <styler2go> i'll try that, thanks a lot! where can ifind those informations if i need them again?
[00:39:37 CEST] <DHE> styler2go: not quite official, but grep gpl configure | grep _deps
[00:40:24 CEST] <styler2go> thank you :)
[00:40:48 CEST] <styler2go> i checked the ffmpeg wiki but there was no notr about gpl
[00:43:07 CEST] <rsully> I have a m2ts file that mediainfo reports as AVC/h264 video, but when I ffprobe it it cannot figure out any info about it (e.g. dimensions) and attempting to mux to mp4 fails. Can I force parameters like width/height to try and get a readable output? I am fairly certain it is not encrypted, though I guess I wouldn't know how to verify
[01:29:57 CEST] <fox_> can anyone kindly help http://stackoverflow.com/questions/37235264/ffmpeg-avcodec-open2-changed-audio-format-in-avcodeccontext
[02:18:46 CEST] <JustinHitla> herp
[02:19:24 CEST] <JustinHitla> derp
[02:27:58 CEST] <koz_> I tried converting some .flac files I have into .opus (with the opus codec) and .ogg (with the libvorbis codec), at 96k bitrate for both. I found that the .oggs were a bit smaller than the .opuses. Is it because .opus has higher quality at the same bitrate?
[02:31:35 CEST] <JustinHitla> koz_: you will not believe it, #opus
[02:31:55 CEST] <koz_> JustinHitla: Are you saying I should have asked there?
[02:33:03 CEST] <JustinHitla> koz_: try
[02:33:49 CEST] <koz_> JustinHitla: OK.
[02:35:23 CEST] <JustinHitla> no, don't leave
[02:38:03 CEST] <fox_> JustinHitla: Can you help me?
[02:38:51 CEST] <JustinHitla> fox_: what ?
[02:39:07 CEST] <JustinHitla> anyone knows why filters are incompatible with stream copying ? any particular reason or just hard to implement ?
[02:39:26 CEST] <fox_> JustinHitla: http://stackoverflow.com/questions/37235264/ffmpeg-avcodec-open2-changed-audio-format-in-avcodeccontext
[02:40:38 CEST] <JustinHitla> fox_: no, I don't know about that
[02:41:06 CEST] <c_14> JustinHitla: filters work on decoded video, therefore you have to decode and then reencode the video
[02:41:38 CEST] <c_14> fox_: try asking on the libav-user@ mailing list
[02:41:49 CEST] <JustinHitla> by filters I mean that "-filter_complex" of course
[02:41:56 CEST] <c_14> yes
[02:45:06 CEST] <fox_> c_14: do i just send email to libav-user at ffmpeg.org ?
[02:45:56 CEST] <c_14> fox_: yes, if you're not subscribed be sure to mention it so that people cc you (or just subscribe)
[02:46:29 CEST] <JustinHitla> in that command:
[02:46:30 CEST] <JustinHitla> ffmpeg -i input1.mp4 -i input2.webm -filter_complex "[0:v:0] [0:a:0] [1:v:0] [1:a:0] concat=n=2:v=1:a=1 [v] [a]" -map "[v]" -map "[a]" -an output.mkv
[02:46:34 CEST] <JustinHitla> is there a way to avoid "-an" ? should I just remove "[0:a:0]" and "[1:a:0]" ?
[02:47:08 CEST] <c_14> remove those, set a=0 and remove the [a] and the -map "[a]"
[02:47:13 CEST] <c_14> or just remove the -map "[a]"
[02:47:25 CEST] <c_14> You're both explicitly mapping the audio stream and explicitly unmapping it
[02:49:48 CEST] <c_14> But removing all references to the audio streams is the "correct" way to do it.
[06:14:28 CEST] <Admin__> hey guys... can you see this http://pastebin.com/raw/wkBMnP7L
[06:14:41 CEST] <Admin__> how do i get around that ? when i run ffproble i see it says stereo and stuff..
[06:14:48 CEST] <Admin__> but i can't seem to get it to pass correctly
[06:43:44 CEST] <ploop> how can I burn in subtitles, and hopefully do so without losing any quality whatsoever from the original video?
[06:58:21 CEST] <mr_lou> I have a noob question about encoding. About bitrates for video. I don't understand why there isn't such a thing as a "optimal-value" parameter that uses the exact bitrate that's necessary.
[06:59:13 CEST] <mr_lou> Like, if I've chosen to have a keyframe every 50 frames, and the whole stream is a 50 fps stream, then the bitrate should logically be "the data for one full frame" + "the differences for the other 49".
[06:59:37 CEST] <mr_lou> Setting a higher bitrate would only result in redundant data, as far as I can see.
[06:59:49 CEST] <mr_lou> And setting a lower bitrate would of course result in a loss of quality.
[07:00:52 CEST] <mr_lou> Or maybe I'm missing something? Maybe there's already a way to get this optimal bitrate for a given video, using some command-liner?
[07:04:46 CEST] <furq> there is no such thing as "the exact bitrate that's necessary" unless you're talking about lossless encoding
[07:05:40 CEST] <furq> lossy encoding is always a tradeoff between quality and size, and perceived quality is for you to decide
[07:06:11 CEST] <furq> but either way you probably want to use crf instead of bitrate-based ratecontrol
[07:06:36 CEST] <furq> unless you're going to be putting the output file in a bunch of 100MB rar files after it's done encoding
[07:42:10 CEST] <Admin__> hey furq .. maybe you know this..
[07:42:16 CEST] <Admin__> http://pastebin.com/raw/wkBMnP7L
[07:42:45 CEST] <Admin__> any way around that.. seems when i run ffproblem it says its aac and stereo and data is there.. soon as i run ffmpeg , it sees channels as 0 and can't copy audio over :(
[07:43:50 CEST] <Admin__> i did raise the " Consider increasing the value for the 'analyzeduration' and 'probesize' options ' ' and still nothing
[07:49:43 CEST] <mr_lou> furq, Your opinion on the socalled Kush Gauge?
[07:59:29 CEST] <furq> depends if my dealer's got any in
[08:05:11 CEST] <furq> ok i googled it and it pretty much looks like nonsense
[08:06:12 CEST] <furq> unless you're streaming or something you should never even consider the bitrate
[08:07:46 CEST] <furq> Admin__: paste the actual ffmpeg command, not ./mysteryscript.sh
[08:10:10 CEST] <Admin__> http://pastebin.com/raw/aw3zaweJ
[08:10:21 CEST] <Admin__> so this is an rtmp source .. that i am getting
[08:10:57 CEST] <Admin__> and ignore fifo.. was from previous and i am using udp  local port now
[08:15:29 CEST] <mr_lou> furq, I'm creating a blu-ray disc. I have a size-limitation of 25 gb for a lot of videos and music. So I need to at least not "waste" bytes.
[08:16:20 CEST] <Admin__> ok and look at this
[08:16:20 CEST] <Admin__> http://pastebin.com/raw/WypbjfXT
[08:17:19 CEST] <Admin__> so command looks sort of like this : http://pastebin.com/raw/aw3zaweJ    & the output  http://pastebin.com/raw/WypbjfXT .... interesting part is.. ffprobe shows me the audio data and ffmpeg does not.. using ffmpeg 3.0.1
[08:22:27 CEST] <furq> Admin__: the default probesize and analyzeduration are 5000000
[08:22:41 CEST] <furq> you probably want to set them higher than that, not lower
[08:27:12 CEST] <Admin__> sure.. let me try but.. i thought i went higher
[08:27:18 CEST] <Admin__> what's a good number ?
[08:31:25 CEST] <Admin__> tried 20M , tried 50M.. tried to increase duration. nothing seems to work
[10:10:16 CEST] <Admin__> nobody knows ?
[13:13:52 CEST] <arunpyasi> Hello guys, I am having an issue. I am unable to install gaspreprocessor though I followed the steps of copying to /usr/local/bin/ and chmod +x to the file.
[14:27:00 CEST] <arunpyasi> Hello guys, I am having an issue. I am unable to install gaspreprocessor though I followed the steps of copying to /usr/local/bin/ and chmod +x to the file. 
[15:25:19 CEST] <spoon> Hi, for H264 decoding, if I decode I-frames in the incorrect order because of network issues. Is it up to me not to decode/display them, or is up to the decoder
[15:33:55 CEST] <DHE> you're responsible for delivering all frames in DTS order. if there are known errors, well...
[15:43:45 CEST] <spoon> ok thanks
[15:55:32 CEST] <DelphiWorld> yo all
[15:55:39 CEST] <DelphiWorld> for stream encoding, its good to use quicksync?
[15:58:54 CEST] <DHE> broadly speaking hardware acceleration speeds up the process (less CPU required) but hurts quality at least a little bit
[16:02:42 CEST] <DelphiWorld> DHE, not looking for a high quality holywood movie, honestly ;)
[16:02:48 CEST] <DelphiWorld> but looking for number of streams
[16:04:56 CEST] <Mavrik> You probably have limited amount of streams the HW encoder can handle
[16:28:48 CEST] <arunpyasi> Hello guys, I am having an issue. I am unable to install gaspreprocessor though I followed the steps of copying to /usr/local/bin/ and chmod +x to the file.
[16:45:19 CEST] <_robert> hello
[16:45:35 CEST] <__jack__> arunpyasi: stop spam
[16:45:42 CEST] <_robert> what
[16:45:45 CEST] <_robert> spam
[16:46:07 CEST] <arunpyasi> I am sorry __jack__
[16:46:28 CEST] <arunpyasi> How can I solve it __jack__  ?
[16:46:54 CEST] <_robert> ok, i'll move to the question: i am trying to convert a .mov to .flv, i used 'ffmpeg -i input.mov -f flv output.flv' but it seems that it is not encoding the alpha channel, so i get a black background, anyone know how could i encode the alpha channel for the flv?
[16:47:08 CEST] <__jack__> arunpyasi: dunno, I've no idea of what's gaspreprocessor
[16:47:20 CEST] <furq> _robert: -c copy
[16:47:32 CEST] <arunpyasi> ok __jack__
[16:47:42 CEST] <_robert> so -f flv -c copy output.flv?
[16:47:46 CEST] <furq> yes
[16:47:52 CEST] <furq> if you need to reencode then specify -pix_fmt yuva420p
[16:47:59 CEST] <furq> but mov to flv will more likely work without reencoding
[16:48:09 CEST] <furq> more than likely
[16:48:25 CEST] <_robert> getting Could not write header for output file #0 (incorrect codec parameters ?): Function not implemented
[16:49:23 CEST] <furq> i guess you do need to reencode then
[16:49:55 CEST] <arunpyasi> furq: can you help me with gaspreprocessor please ?
[16:50:28 CEST] <furq> i don't know what that is
[16:52:19 CEST] <_robert> hmm @furq the -pix_fmt yuva420p is not encoding the alpha channel, i still get the black background :(
[16:54:30 CEST] <_robert> https://codeshare.io/56IRM
[16:55:33 CEST] <furq> Incompatible pixel format 'yuva420p' for codec 'flv', auto-selecting format 'yuv420p'
[16:55:36 CEST] <furq> fun
[16:55:45 CEST] <_robert> oh, yuva the a was for alpha?
[16:55:52 CEST] <_robert> and it did a fallback on yuv
[16:55:53 CEST] <furq> i forgot the auto codec selection is flv1 for flv
[16:56:01 CEST] <furq> try specifying -c:v libx264
[16:56:21 CEST] <furq> i've never seen an h.264 video with alpha, though, so i have no idea if it'll work or be supported by players
[16:56:23 CEST] <_robert> should i delete the -f flv or just add the -c
[16:56:30 CEST] <furq> you can delete -f flv, it's redundant
[16:57:20 CEST] <kepstin> i think if you want alpha in an flv file, you need to use one of the vp6 codec varients
[16:57:42 CEST] <kepstin> 'vp6a' specifically
[16:57:50 CEST] <_robert> well, i guess i am just unlucky
[16:58:09 CEST] <furq> does it have to be flv
[16:58:16 CEST] <_robert> yes, i need to embed it in flash
[16:58:31 CEST] <furq> does it have to be flash or just anything you can view in a browser
[16:58:43 CEST] <furq> if it's the latter then webm definitely supports an alpha channel
[16:59:10 CEST] <_robert> unfortunatelly it needs to be in the old dieing flash
[16:59:40 CEST] <furq> looks like you're stuck with vp6 then
[17:00:31 CEST] <_robert> how do i use that? with -c? is for codec?
[17:00:38 CEST] <furq> presumably -c:v vp6a
[17:00:42 CEST] <_robert> ty
[17:00:57 CEST] <kepstin> looks like ffmpeg doesn't support encoding vp6, only decoding it
[17:01:04 CEST] <furq> oh nice
[17:01:06 CEST] <_robert> oh (-c = codec : v = video) right?
[17:01:10 CEST] <furq> yeah
[17:01:18 CEST] <kepstin> you'll probably need to get a commercial encoder :/
[17:01:54 CEST] <_robert> actually i installed an older version of adobe media encoder and did it already, but wanted to find out for the future if i could use the ffmpeg
[17:02:04 CEST] <_robert> the removed the flv export from cc 2014 and newer
[17:02:07 CEST] <_robert> they*
[17:03:31 CEST] <kepstin> oh, fun, since google bought ON2 they actually stopped selling the commercial VP6 encoder :)
[17:17:32 CEST] <DelphiWorld> sup
[17:17:55 CEST] <DelphiWorld> DHE, any recomanded intel I7 with quicksync?
[17:19:31 CEST] <DHE> no idea. I don't use have
[17:19:53 CEST] <DelphiWorld> have? :P
[17:19:59 CEST] <DelphiWorld> hardware assisted video encoding?
[17:20:38 CEST] <DHE> nvenc, opencl, black-box vendor stuff
[17:21:02 CEST] <DelphiWorld> yeah
[17:21:24 CEST] <DHE> I dno't use or have it
[17:21:31 CEST] Action: DHE is incoherent
[17:21:46 CEST] <DelphiWorld> oki;)
[17:22:29 CEST] <kepstin> i'd expect the qsv block to be basically identical on any intel chip of the same generation
[17:22:43 CEST] <DelphiWorld> DHE: i have a E5 6 cores... it get load with less thant 11 sd streams
[17:22:47 CEST] <vade> if I run audio packets through a filter chain via libfilter do I need to retime them via av_scale_q if my audio buffer and my audio sync have different sample rates ?
[17:23:34 CEST] <DelphiWorld> so i thought HAVE should be usefull
[17:24:58 CEST] <kepstin> hmm. apparently the qsv speed does scale with the gpu clock speed, so going for a processor with higher gpu clocks might get you slightly better qsv performance.
[17:25:42 CEST] <DHE> vade: you should be able to specify that you want different input and output settings. the libavfilter chain will automatically insert a resampler for you to clear up the difference
[17:26:00 CEST] <DelphiWorld> kepstin: good info
[17:26:20 CEST] <vade> yup, im doing that - and it works perfectly. however im getting a difference in duration from my input audio track and my output audio track, and im trying to track down the source
[17:26:41 CEST] <vade> ie: does the avframe* that I pass into the filter need its duration changed to match the sample rate conversion happening internally to the filter?
[17:30:49 CEST] <jkqxz> DelphiWorld:  The Iris Pro ones should be significantly faster, because they have doubled codec blocks.  Intel definitely does funny stuff, though, so you probably want to actually test whatever setup you're aiming for.
[17:31:31 CEST] <DelphiWorld> jkqxz: any recomandation? :P
[17:31:34 CEST] <jkqxz> Still, with a desktop chip you have easily got >200fps 1080p H.264 transcode.  That's a lot of SD.
[17:31:41 CEST] <DelphiWorld> jkqxz: it'll be used to encode iptv streams
[17:34:50 CEST] <jkqxz> My one data point for funny stuff happening is that a Skylake 6300 (desktop) is about half as fast as a Haswell 4500U (mobile), despite the specs suggesting it should be significantly faster.
[17:35:25 CEST] <DHE> meh, I just throw e5-2690+ CPUs at it. :)
[17:35:31 CEST] <jkqxz> That pretty much destroys any confidence I have in extrapolation of performance from specifications, so I'm left with "test it".
[17:35:35 CEST] <DHE> nvenc performs admirably too
[17:38:42 CEST] <kepstin> i've always been a bit annoyed how intel puts much better gpus on their mobile chips than desktop chips, but I guess that's where the money is :/
[17:39:37 CEST] <jkqxz> If you go by specifications, then either a 67??HG (in Skull Canyon NUC, release about now?) or an E3-15?5M (???) /should/ be the fastest platform you can get.  Whether it actually is is entirely unclear.
[17:42:37 CEST] Action: kepstin is actually kind of tempted to pick up an i7-577C for one of his boxes, but it's ridiculously expensive :/
[17:43:20 CEST] <kepstin> 5775C
[17:43:40 CEST] <lavalike> hello, how can I work around this problem: [libx264 @ 0x2a6bf40] height not divisible by 4 (1200x674) --- I am re-encoding H.264 video
[17:45:56 CEST] <furq> lavalike: use a height which is divisible by 4
[17:46:02 CEST] <furq> or deinterlace the source
[17:46:14 CEST] <lavalike> furq: so this would add black bars at the top and bottom, or truncate some pixels?
[17:46:35 CEST] <furq> that's one way
[17:46:42 CEST] <lavalike> mediainfo says "Progressive" for the video stream, if I understand what you mean by interlacing
[17:46:58 CEST] <furq> afaik it should only complain about non-mod4 height for interlaced video
[17:47:05 CEST] <lavalike> uhm
[17:47:36 CEST] <lavalike> http://pastebin.com/raw/jKUYfF8j
[17:47:50 CEST] <vade> yea im at a loss about this audio timing. its never dead on like my video is
[17:48:00 CEST] <vade> (using libswrescale)
[17:48:18 CEST] <furq> lavalike: yeah that's weird
[17:48:32 CEST] <furq> obviously the source is h.264 and has a non-mod4 height
[17:48:56 CEST] <furq> assuming that's an error and not a warning, i can't see why x264 would be complaining about that now
[17:48:59 CEST] <lavalike> and this is ffmpeg's invocation: http://pastebin.com/raw/b8FFdEqW
[17:49:07 CEST] <lavalike> it stumped me as well
[17:49:26 CEST] <furq> why are you passing -flags +ildtc
[17:49:29 CEST] <furq> ct
[17:49:34 CEST] <kepstin> you're using "-flags +ildct" which tells x264 to do an interlaced encode...
[17:49:36 CEST] <furq> yeah
[17:49:40 CEST] <lavalike> oh god
[17:50:04 CEST] <lavalike> Scan type: MBAFF
[17:50:09 CEST] <lavalike> you're right, what have I done
[17:51:36 CEST] <furq> i also wonder why you're using -tune stillimage for what appears to be a film
[17:51:42 CEST] <furq> but maybe it is a 90-minute still image
[17:51:55 CEST] <lavalike> yes, it basically is
[17:52:15 CEST] <furq> i guess that explains why you're reencoding it then
[17:52:20 CEST] <lavalike> it is a very static computer screen recording, exactly
[17:52:34 CEST] <lavalike> man I have encoded so much stuff in interlaced form by accident apparently
[17:52:53 CEST] <kepstin> the 'stillimage' tune is designed for encoding a single frame as an image format, it's unclear what effect it has on a multi-frame video :/
[17:53:24 CEST] <kepstin> you might have better results with something like '-tune animation' or '-tune touhou' for a screen capture with only a small amount of movement.
[17:53:32 CEST] <kepstin> have to test and see.
[17:53:42 CEST] <lavalike> I don't think I tried those two when I started the project
[17:53:48 CEST] <furq> why did they have to call it "touhou" instead of a good game
[17:53:55 CEST] <lavalike> that made me laugh
[17:54:33 CEST] <lavalike> is the interlaced <-> progressive transformation lossless?
[17:55:03 CEST] <lavalike> .oO("not even remotely")
[17:55:20 CEST] <kepstin> lavalike: encoding progressive content with the interlaced encoding enabled in x264 is a mild efficiency loss because it needs an extra flag per macroblock, iirc
[17:55:31 CEST] <kepstin> it doesn't actually "transform" it to lossless
[17:55:39 CEST] <kepstin> er, "transform" it to interlaced i mean
[17:56:00 CEST] <lavalike> kepstin: ah interesting, I was pondering transforming them back, over a period of time,
[17:57:33 CEST] <kepstin> basically, the way the mbaff mode used in x264 works is that if it thinks a macroblock looks interlaced, it'll do a lossless transform to put all the even lines in the top half and odd lines in tbe bottom half before doing the encode; this makes the rest of the encoder slightly more efficient at encoding that block
[17:57:54 CEST] <kepstin> the downside is that it adds 1 bit per macroblock to say whether this transform is applied or not on that block
[17:58:34 CEST] <lavalike> super clean explanation, cool, I think I will not transform them back, but stop adding unnecessary bits from now on
[17:58:44 CEST] <kepstin> so yeah, no reason to re-encode, you'll only lose more quality due to a generation loss on the re-encode.
[18:00:02 CEST] <lavalike> I think I'll try both of the -tune settings you suggested and see if they perform better
[18:01:23 CEST] <lavalike> when I first looked into parameters' values I extracted small pieces of video and tried encoded them with various crf values and various presets, compared, plotted, and decided on one: do you think I need to go through this again for these two or are the crf values comparable and I can only look at overall size?
[18:02:57 CEST] <kepstin> hmm. those tunes both affect the deblocker, iirc, so they can give a different "look" to the video
[18:06:10 CEST] <kepstin> both animation and touhou bump up the allowed reference frames significantly, which should reduce filesize a bit on mostly-static video.
[18:06:58 CEST] <kepstin> (stillimage doesn't do that, because it's designed for only encoding 1 frame, so there are no predicted frames anyways)
[18:07:31 CEST] <lavalike> to be extra-precise they are, I think, more than 50% static overall and the changes are always pretty much local to small rectangular subsections of the frame, like windows on screen
[18:08:20 CEST] <lavalike> usually there's a face in a corner moving very little and talking, so I tuned the parameters looking at blocking in the face features mostly
[18:08:51 CEST] <kepstin> yeah, in that case having more reference frames can help, because it'll be able to spend fewer bits on the static section of the video.
[18:08:56 CEST] <kepstin> maybe
[18:09:05 CEST] <kepstin> have to try and see I guess :/
[18:09:11 CEST] <lavalike> alright :)
[18:10:14 CEST] <lavalike> I wish I would have caught this earlier, I would have tested over the weekend, oh well
[18:10:21 CEST] <lavalike> thank you for the info, much appreciated
[18:12:27 CEST] <kepstin> but yeah, main difference between animation and touhou is that animation bumps up the deblocker to make flat color areas look smooth, touhou reduces it to keep the appearance of more detail, and there's some funkiness in the touhou one where the psy tuning is adjusted to make small moving things keep detail better (this might affect e.g. mouse cursors)
[18:13:29 CEST] Action: kepstin wonders what combination of settings would be ideal for a 'screencapture' tuning mode, maybe somewhere inbetween the two?
[18:14:40 CEST] <lavalike> nice, I think touhou is probably going to work best
[18:15:58 CEST] <lavalike> there is "grain" and texture in a majority of the frame, as well as translucent/semi-transparent objects moving/changing every once in a while
[18:16:37 CEST] <lavalike> whereas the smoothing sounds counter-productive for a screencapture in my mind
[20:29:05 CEST] <Helle> so erm, perhaps a silly question and I have /a/ answer, but I am wondering if it is possible to do more efficiently, I want to extract a thumbnail from a video stream every n-seconds, without having ffmpeg try and decode all of it (just decode the I-frame closest to the nth second and save that would be fine), current solution is to just have ffmpeg fire up every n secondss and dump a single frame, then exit, but this is a bit inefficient due to the fact it 
[20:32:14 CEST] <kepstin> Helle: The '-discard nokey' input option might work on some types of streams, you could give it a try.
[20:32:34 CEST] <kepstin> it would just discard all non-keyframes, rather than all except the ones at intervals you prefer
[20:33:17 CEST] <Helle> kepstin: still a significant performance boost
[20:33:32 CEST] <Helle> kepstin: but nothing closer to what I want ?
[20:33:45 CEST] <kepstin> Helle: to do exactly what you want, I think you'd have to make a custom app using libavformat/libavcodec
[20:33:53 CEST] <kepstin> it's not really a common use case :)
[20:33:56 CEST] <Helle> kepstin: true :)
[20:34:30 CEST] <Helle> I guess it might be doable by taking a look at how nokey is implemented and making a custom program that just skips a little more
[20:35:48 CEST] <kepstin> I think avdiscard is actually implemented in the demuxer and/or decoder, so it wouldn't be easy to change
[20:35:59 CEST] <Helle> kepstin: oh, *pout*
[20:36:44 CEST] <kepstin> if you did a custom app, I think the process would be: read data, demux it, inspect the frame timestamps and whether it's a keyframe to decide whether to forward it to decoder or drop it on the floor.
[20:37:57 CEST] <Helle> mhm, might be worth looking at I guess, although this is about 10x more efficient already, neat
[20:39:41 CEST] <Helle> feels so weird to be this deep into media, haven't poked around this level since uuuuh, when you had to manually specify the I/P/B frame patterns for your mpeg 1 encoder
[20:40:03 CEST] <Helle> usually at a slightly higher level or on lossless compression which uses such different methods
[20:44:11 CEST] <Helle> packet.flags & AV_PKT_FLAG_KEY welp, looks simple enough, and I bet the codec would be more then happy to just decode I-frames
[20:47:48 CEST] <Helle> kepstin: cool, I think I might be able to just code this
[20:53:22 CEST] <Helle> (presuming the codecs are fine just being fed I-frames)
[20:59:28 CEST] <sasha> Is there a way I can fake a moov atom with ffmpeg?
[20:59:55 CEST] <sasha> I'm trying to convert a bunch of Seg1-Frag1 files into a readable format from an Adobe HDS stream, it's really annoying me
[21:01:36 CEST] <sasha> here is an example file https://foo.fixme.ch/public/1_3229ca0a8e7e7e34_Seg1-Frag1
[21:01:58 CEST] <sasha> I can see some information in the plaintext at the top, but I cannot figure out how to get it to play with anything (vlc, mplayer..) and ffmpeg doesn't open it
[21:02:00 CEST] <sasha> any ideas?
[21:05:15 CEST] <lesshaste> how can I reduce the fps in a video and keep everything else the same?
[21:38:20 CEST] <kepstin> lesshaste: not sure what you mean. Slow the video down? or drop some frames so motion is more jerky?
[22:07:30 CEST] <pgorley> When I compile on OS X 10.10, the ffmpeg tool can't find CoreImage.framework
[22:07:47 CEST] <pgorley> I'm doing the normal .'configure followed by make && sudo make install
[22:08:10 CEST] <pgorley> * ./configure
[22:09:36 CEST] <pgorley> So I found that the CoreImage framework is also underneath QuartzCore framework, so I copied it to the /System/Library/Frameworks folder
[22:10:39 CEST] <pgorley> Now it exits complaining that the symbol _kCVImageBufferColorPrimaries_ITU_R_2020 is not found in the CoreVideo framework
[22:11:38 CEST] <pgorley> I'm running Yosemite
[23:57:37 CEST] <vikash>  http://pastebin.com/ZPPLEARk
[23:57:55 CEST] <vikash> Can anybody tell me the reason for above warning ?
[00:00:00 CEST] --- Tue May 17 2016