[Ffmpeg-devel-irc] ffmpeg.log.20161018

[burek]

[00:37:09 CEST] <Phi> The libx264 still isn't taking any options
[00:37:58 CEST] <c_14> Try convincing it that if it invest now the stock price will only rise in the future.
[00:38:35 CEST] <c_14> (I'm going to need more information if I'm going to try and help you)
[00:40:22 CEST] <Phi> mm
[00:40:34 CEST] <Phi> but I thought everyone here was psychic!
[00:40:50 CEST] <c_14> Not on Monday.s
[00:41:12 CEST] <Phi> fair enough
[00:41:14 CEST] <Phi> http://stackoverflow.com/questions/40053873/ffmpeg-rtsp-stream-to-mpeg4-h264-file-using-libx264
[00:41:36 CEST] <Phi> basically, using AVFormatContext to chuck a H264 stream into a MPEG4 container, using libx264 encoder
[00:41:50 CEST] <Phi> I can't change any of the options, the profile is always High
[00:42:14 CEST] <Phi> av_opt_set(video_file_codec_context->priv_data, "profile", "main", AV_OPT_SEARCH_CHILDREN);
[00:42:20 CEST] <Phi> seems to be the main suggestion
[00:43:52 CEST] <Phi> I've checked the examples and I've even proxied libx264 Codec's Init function pointers, but they're not even called
[00:46:56 CEST] <Phi> what's weirder is I check the X264Context, and the encoder variable isn't even set
[00:47:07 CEST] <Phi> yet a mp4 file is produced, which is very confusing
[00:51:35 CEST] <Phi> any idea c_14?
[00:56:03 CEST] <c_14> If anything, you should be calling av_opt_set directly before marker3
[00:59:32 CEST] <Phi> should I be calling avcodec_open2?
[00:59:35 CEST] <Phi> it doesn't seem used
[00:59:56 CEST] <c_14> It's in the examples, so probably
[01:00:34 CEST] <c_14> https://git.ffmpeg.org/gitweb/ffmpeg.git/blob_plain/HEAD:/doc/examples/decoding_encoding.c
[01:08:44 CEST] <Phi> I get the feeling I'm doing something weird
[01:10:00 CEST] <Phi> because I use AVIOContext, AVFormatContext, AVCodecContext...
[01:10:18 CEST] <Phi> that example isn't very accurate, it's just input -> images, which I can do
[01:10:44 CEST] <Phi> I need output video file
[01:11:17 CEST] <Phi> I mostly pulled my code from http://stackoverflow.com/questions/9251747/record-rtsp-stream-with-ffmpeg-libavformat
[01:47:05 CEST] <blue_misfit> I'm seeing something crazy.. if I made an H.264 proxy with no b-frames of an editing codec like ProRes or DNxHD then I can drop it on top of the source in Premiere or Resolve and it's absolutely perfect... but if I enable b-frames it's immediately out of sync
[01:47:10 CEST] <blue_misfit> has anyone else observed this??
[01:48:37 CEST] <DHE> graphical glitches, or just audio drift?
[01:49:40 CEST] <blue_misfit> immediate audio sync offset equal to the number of bframes
[01:49:50 CEST] <blue_misfit> sorry not audio sync offset
[01:51:17 CEST] <blue_misfit> the version with b-frames is slipped back by a few frames
[01:51:27 CEST] <blue_misfit> equal to the number of consecutive b-frames the proxy is encoded with
[01:56:21 CEST] <blue_misfit> so like if a video normally starts at 00:00:00:00
[01:56:26 CEST] <blue_misfit> it will start at 00:00:00:02
[01:56:34 CEST] <blue_misfit> if you encode with 3 consecutive b-frames
[01:58:03 CEST] <blue_misfit> hmm actually looks like it's just always a 2 frame offset.. using 3 bframes or 10 bframes always shows the same result
[02:09:02 CEST] <lindylex> Can I get ffmpeg to get the title of the youtube video file and not download the video?
[02:10:07 CEST] <furq> lindylex: use youtube-dl
[02:10:21 CEST] <lindylex> Thanks
[02:10:25 CEST] <furq> http://vpaste.net/6IHoZ
[02:10:48 CEST] <lindylex> furq: you are the best!
[02:15:45 CEST] <blue_misfit> DHE, looks like it actually works with 1 b-frame but any more than that and you immediately have a 2 frame offset in Premiere
[02:15:51 CEST] <blue_misfit> very odd. I'm chatting with their support about this
[02:19:20 CEST] <DHE> b-frames are decoded and displayed out of order. poor players or buggy muxers may screw it up resulting in that...
[02:19:34 CEST] <DHE> but I don't know what app you're using so I'm just going to stop here
[02:20:00 CEST] <furq> he said it's premiere
[02:20:07 CEST] <blue_misfit> yeah - Premeiere Pro 2015.4
[02:20:09 CEST] <blue_misfit> it's fun stuff
[02:20:25 CEST] <furq> doesn't premiere use the system (dshow/quicktime) codecs
[02:20:26 CEST] <DHE> still havn't used it...
[02:20:37 CEST] <blue_misfit> maybe - it also might use its own internal codecs
[02:20:59 CEST] <blue_misfit> per support "I believe this is by Design with THE GOP size(in frames) , Its a general rule that when I frames is set to 1 then all frames are I frmaes, but when its largesr than 1 its then when the action takes place. That is why its coming as 00:00:00:02"
[02:21:00 CEST] <blue_misfit> LOL
[02:21:26 CEST] <furq> what about p frames
[02:21:34 CEST] <furq> or frmaes, rather
[02:22:11 CEST] <DHE> uhhh...
[02:22:39 CEST] <DHE> while all I-frames are technically valid and recognized as a profile, nobody does that except for special situations
[02:22:54 CEST] <furq> yeah no bframes has nothing to do with gop size
[02:24:11 CEST] <furq> if it's happening in resolve as well then i'd suspect it's an issue with the system decoders
[02:24:21 CEST] <furq> which i'm sure will be fun to, uh, resolve
[02:24:56 CEST] <DHE> I would guess something is reading the frames one at a time from the file, and presenting them one at a time as they come out of the decoder without really understanding the need to buffer them..
[02:25:14 CEST] <blue_misfit> yeah agreed
[02:25:26 CEST] <blue_misfit> this is crazy though! It's Premiere!! It's widely known and trusted for editing video!
[02:25:49 CEST] <DHE> still, b-frames are business as usual for H1264 and h264 has been around for a while now. this is a rookie sort of error...
[02:26:02 CEST] <blue_misfit> of course
[02:26:23 CEST] <blue_misfit> there's definitely nothing wrong with these files - decoding them back to y4m returns the same number of bytes so it's not like frames are being lost
[02:26:28 CEST] <blue_misfit> they're also totally there in Elecard
[02:52:50 CEST] <Phi> I might join later, cheers for your help c_14
[03:35:54 CEST] <beauty> Is there a avformat_copy_context funtion like "avcodec_copy_context"
[03:37:54 CEST] <DHE> like avcodec_parameters_from_context ?
[03:38:07 CEST] <DHE> part of the codecpar system
[03:38:17 CEST] <k_sze[work]> When I use av_interleaved_write_frame, is any of the errors recoverable?
[03:42:02 CEST] <beauty> DHE: year
[03:42:35 CEST] <beauty> I want to store the state of avformatcontext
[03:43:20 CEST] <DHE> ... do you mean save the state of the encoder?
[03:43:52 CEST] <beauty> year
[03:43:55 CEST] <beauty> decoder
[03:44:23 CEST] <DHE> year?
[03:45:08 CEST] <beauty> I want to open ffmpeg,decode ,close ffmpeg. Then open again,av_seek to the last position, decode ,close ffmpeg. repeat
[03:46:11 CEST] <DHE> that's not quite how it works. internal codec.... stuff... isn't just in the avcodeccontext. this strategy could work only if you make sure to bail on a keyframe or are ready to deal with exact seeking without keyframes.
[03:46:50 CEST] <DHE> but using that method, saving the avcodeccontext isn't going to help. you're saving the frame number or exact timestamp, etc
[03:48:00 CEST] <beauty> so,how to realise my requirement?
[03:50:22 CEST] <beauty> I want use ffmpeg to decode 4000 streamed videos.
[03:55:54 CEST] <beauty> how to use av_probe_input_format get avinputformat?
[06:03:10 CEST] <Phantom_X5> hello, anyone interested in helping me?
[06:03:39 CEST] <qmr> I'm interested in you not wasting our disk space with useless questions
[06:03:50 CEST] <qmr> Do not ask to ask, just ask.  If someone is willing and able to help, they will
[06:04:13 CEST] <qmr> the other big thing is ask then WAIT.  so many people join, ask, leave in 2 minutes when they don't get immediate help
[06:06:08 CEST] <Phantom_X5> I have an issue with ffmpeg moving an output file to a different folder on the webspace. File uploads, moves to temp folder, ffmpeg converts successfully, file does not move to completed folder. Not sure if it is a permissions problem or something else
[06:08:47 CEST] <qmr> that's all out of the context of ffmpeg as far as I know
[06:08:55 CEST] <qmr> what web app are you using?
[06:09:49 CEST] <qmr> I was often considered the smartest admin in the room at one of my old jobs.  really I'm old and senile and confused most of the time, however, you can solve pretty much all problems with ...........
[06:10:39 CEST] <qmr> logs, verbose output, google, documentation, strace/dtrace, tcpdump, source code
[06:11:02 CEST] <qmr> in particular you should have logs of some kind for whatever web thing you are running
[06:14:27 CEST] <Phantom_X5> I'm starting out with testing clipbucket
[06:15:55 CEST] <Zeranoe> Phantom_X5: A bit of your server side code would go a long way.
[06:17:41 CEST] <Phantom_X5> can you be more specific?
[06:18:17 CEST] <Zeranoe> Phantom_X5: How are you calling FFmpeg
[06:18:44 CEST] <Phantom_X5> not sure.. I'll see what I can find
[06:22:57 CEST] <Phantom_X5> there is a ffmpeg.class.php file
[06:27:52 CEST] <Phantom_X5> I will have to dig through this to find exactly what it is asking ffmpeg to do. It might be asking apache to deal with the finished file instead of ffmpeg. If I find the source, I'll come back
[06:27:57 CEST] <Phantom_X5> thanks for the help tho
[10:48:01 CEST] <i11ega1Code> hi i have a problem: https://pastebin.mozilla.org/8919776
[11:09:18 CEST] <i11ega1Code> how do i make mp4 playable for psp on ubuntu? i need something like oneline command that will convert the files of a directory and delete them when converted
[11:20:05 CEST] <hero_biz> hi guys.
[11:21:09 CEST] <JEEB> i11ega1Code: http://x264.fushizen.eu/random/psp_encoding.txt
[11:21:34 CEST] <JEEB> (I should really update that to x264-params instead of x264opts
[11:21:49 CEST] <JEEB> (that file hasn't been updated in like five years by now)
[11:21:59 CEST] <hero_biz> guys, I'm trying to sharp a bad blurry source a little while encoding with ffmpeg. I'm using unsharp=3:3:1.5 for this reason. I wonder if there is any better sharpner, or better params for unsharp.
[11:24:09 CEST] <hero_biz> what happens is I increase luma_msize_x and luma_msize_y?
[11:30:00 CEST] <jubalh> Hi, I am thinking about converting all my avi + prop. codec into ogg theora format. Is there anything that speaks against this? I dont see a reason but it seems like most people dont do this, so I wonder if I am missing something.
[11:31:13 CEST] <JEEB> a) do you have requirements of a colorspace that is not 4:2:0 YCbCr b) do you understand the limitations of theora (eats bits for breakfast)
[11:32:10 CEST] <jubalh> b) no i dont. I read that it should be a major alternative to the prop codecs. I hoped it can help me to not have to install all the prop codecs and thus not add those inofficial repos
[11:32:43 CEST] <jubalh> eating bits sounds like i will miss something regarding quality?
[11:32:55 CEST] <JEEB> uhh, many distros have normal lavc in their normal repos? just maybe not installed by default
[11:33:00 CEST] <JEEB> (like debian/ubuntu)
[11:33:54 CEST] <hero_biz> here is a small snapshot from source: http://oi64.tinypic.com/2qby5pg.jpg which is a disaster quality. is there a better option than unsharp=3:3:1.5 for sharpening this video?
[11:35:01 CEST] <jubalh> JEEB: I am not so familiar with these video things. i wonder if there is something to read which will make me better understand the options i have and the consequences it will have to change to another format
[11:35:20 CEST] <jubalh> so far i totally rely on opinion on others via google.
[11:42:27 CEST] <i11ega1Code> ehm...
[11:42:30 CEST] <i11ega1Code> are you guys here
[11:42:35 CEST] <i11ega1Code> why aren't you answering
[11:43:03 CEST] <hero_biz> i11ega1Code: JEEB gave your answer.
[11:43:24 CEST] <i11ega1Code> i don't see it, perhaps i have disconnected ?
[11:43:37 CEST] <i11ega1Code> could you give me the answer he gave me hero_biz?
[11:44:57 CEST] <i11ega1Code> hero_biz
[11:45:46 CEST] <hero_biz> <JEEB> i11ega1Code: http://x264.fushizen.eu/random/psp_encoding.txt
[11:45:55 CEST] <hero_biz> <JEEB> (I should really update that to x264-params instead of x264opts
[11:46:05 CEST] <hero_biz> <JEEB> (that file hasn't been updated in like five years by now)
[11:50:06 CEST] <i11ega1Code> ok how do i update?
[11:50:45 CEST] <hero_biz> it says, use x264-params rather than x264opts in command line.
[11:51:20 CEST] <i11ega1Code> ok is there something wrong with this command:
[11:51:21 CEST] <i11ega1Code> for file in *.mp4; do outfile=CONVERTED/$file; ffmpeg -y -i "$file" -flags +bitexact -s 400x192 -r 29.97 -b:v 512k -c:a aac -b 672k -ab 96k -ar 24000 -f psp -strict -2 "$outfile"; done
[11:51:51 CEST] <i11ega1Code> i tried that ^ though i need to use:
[11:51:51 CEST] <i11ega1Code> for file in *.mp4; do outfile=CONVERTED/$file; ffmpeg -y -i "$file" -flags +bitexact -s 400x192 -r 29.97 -b:v 512k -acodec libfaac -b 672k -ab 96k -ar 24000 -f psp -strict -2 "$outfile"; done
[11:52:52 CEST] <i11ega1Code> if anyone could help, it would be awesome, i need to go to a trip soon and i would love to watch some episodes on my psp
[11:55:54 CEST] Action: hero_biz smartphones are much more great for watching episodes, I personally use galast s 5 with mx player :)
[11:57:13 CEST] <jubalh> I am reading the ffmpeg manpage and ocumentation. in the command at http://ffmpeg.org/ffmpeg-all.html#Audible-AAX i dont see what '-c:a copy' does?
[11:58:28 CEST] <i11ega1Code> for file in *.mp4; do outfile=CONVERTED/$file; ffmpeg -i "$file" -b 300 -s 320x240 -vcodec xvid -^C 32 -ar 24000 -acodec aac "$outfile"; done
[11:58:28 CEST] <i11ega1Code>  gives me : Unknown encoder 'xvid'
[11:58:48 CEST] <iamtakingiteasy> hi, one wild question regarding mp4 container with aac audio payload in mdat boxes; should aac frames be raw or prefixed by ADTS-header?
[11:59:38 CEST] <hero_biz> jubalh: means copy audio untouched.
[12:00:07 CEST] <hero_biz> i11ega1Code: means your ffmpeg is not build with xvid support.
[12:03:02 CEST] <i11ega1Code> i needed to change it to libxvid encoder
[12:03:07 CEST] <i11ega1Code> :)
[12:04:21 CEST] <jubalh> hero_biz: I see. Is there any pro in doing this to an aax file? i thought about converting to ogg vorbis. But maybe this will make me loose the chapter feature?
[12:06:29 CEST] <fling> How to convert iso to flac?
[12:06:31 CEST] <hero_biz> that option is not related to chapter (if you mean mkv chapter) as i know.
[12:06:34 CEST] <i11ega1Code> YAY WORKED
[12:06:34 CEST] <i11ega1Code> for file in *.mp4; do outfile=CONVERTED/$file; ffmpeg -i "$file" -b 300 -s 320x240 -vcodec libxvid -ab 32 -ar 24000 -acodec aac -strict -2 "$outfile"; don
[12:07:14 CEST] <hero_biz> i'n not expert in encoding though.
[12:07:21 CEST] <jubalh> okay
[12:07:23 CEST] <jubalh> thanks
[12:09:13 CEST] <oleg_mo> hello all, i got problem, i try to install this app http://pspvc.sourceforge.net/ from source, during instalation i got error
[12:09:14 CEST] <oleg_mo> ERROR: XviD not found
[12:09:14 CEST] <oleg_mo> If you think configure made a mistake, make sure you are using the latest
[12:09:14 CEST] <oleg_mo> version from SVN.  If the latest version fails, report the problem to the
[12:09:14 CEST] <oleg_mo> ffmpeg-devel at mplayerhq.hu mailing list or IRC #ffmpeg on irc.freenode.net.
[12:09:14 CEST] <oleg_mo> Include the log file "config.err" produced by configure as this will help
[12:09:14 CEST] <oleg_mo> solving the problem.
[12:09:15 CEST] <oleg_mo> -e \E[01;31mERROR during configure FFMPEG
[12:12:15 CEST] <oleg_mo> sorry, i fix the problem https://help.ubuntu.com/community/PSP.
[12:13:14 CEST] <nonex86> did i miss something? today the day of psp?
[12:17:23 CEST] <oleg_mo> :)
[12:17:43 CEST] <oleg_mo> i just have miss the lib to install
[12:18:20 CEST] <oleg_mo> https://gist.github.com/xdamman/e4f713c8cd1a389a5917#file-install_ffmpeg_ubuntu-sh
[12:18:20 CEST] <oleg_mo> i reinstall with this sh all deps libs
[12:18:20 CEST] <nonex86> check the config.log in ffmpeg folder
[12:18:46 CEST] <nonex86> last records give you an idea why ffmpeg failed to build
[12:19:02 CEST] <nonex86> usually, yes, its some dependencies
[12:36:28 CEST] <fling> How to convert from bin/cue?
[12:40:58 CEST] <fling> ok it is bchunk
[12:45:06 CEST] <ozette> vcodex.com is an interesting website
[12:45:23 CEST] <ozette> found it recently
[12:46:37 CEST] <jubalh> do i see it right that mp4, m4a and m4b are all the same container format. the only difference is in how players treat them? but in theory i could just change filename and there is no conversion needed?
[13:41:56 CEST] <acidflash> Hi
[13:42:06 CEST] <acidflash> Can we discuss ffserver related issues here aswell ?
[13:51:29 CEST] <acidflash> I was wondering if the ffm format is able to "receive" more than just audio and video streams, ie: a subtitle stream
[13:52:40 CEST] <hseg> Hi. I have a ~2h long mkv file. I need to remove the part 01:02:00.5-01:02:02. Is there a way to do this without a massive reencoding job?
[13:53:12 CEST] <hseg> (i.e. I need 00:00:00-01:02:00.5 followed by 01:02:02-the end)
[13:53:19 CEST] <nonex86> hseg: what video codec?
[13:53:52 CEST] <hseg> h264+vorbis
[13:54:23 CEST] <nonex86> well, not sure about ffmpeg itself, but h264 can easily be cut from idr to idr
[13:54:39 CEST] <hseg> Oh?
[13:55:34 CEST] <nonex86> also usually its possible cut at frame basis with some reencoding 1st and last gop's
[13:55:37 CEST] <hseg> (Reason this happened - I concatenated a bunch of episodes, and typo'd the end timestamp for one of them. It is really jarring)
[13:55:50 CEST] <hseg> OK - how do I do it?
[13:56:02 CEST] <nonex86> i guess ffmpeg can cut at idr basis (key frames)
[13:56:40 CEST] <hseg> Not sure the keyframes will align with the gap I want.
[13:57:18 CEST] <nonex86> -ss?
[13:57:35 CEST] <hseg> You're thinking of the inverse problem I'm trying to solve.
[13:57:49 CEST] <hseg> I need to remove a segment of the file, not extract it.
[13:58:09 CEST] <nonex86> :)
[13:58:18 CEST] <hseg> Yeah.
[13:58:35 CEST] <hseg> It would suffice for me if I could blank out that segment.
[13:58:49 CEST] <nonex86> actually this dosnt change the solution :) under the hood
[13:58:58 CEST] <hseg> i.e. replace the frames there by black screen with null audio.
[13:59:04 CEST] <hseg> What do you mean?
[13:59:39 CEST] <nonex86> nevermind, you just asking a functinality i created in my previouse project :)
[14:00:02 CEST] <nonex86> cut/merge h264 streams minimizing stream reencoding
[14:00:25 CEST] <nonex86> so i am looking to your problem from developer perspective :)
[14:01:20 CEST] <hseg> ... Does vorbis also support no-reencoding cuts?
[14:01:40 CEST] <hseg> Because otherwise this is a no-go, as my mkv has h264+vorbis.
[14:02:24 CEST] <nonex86> not sure about vorbis
[14:02:25 CEST] <hseg> I *could* just  reencode everything, it just would take ~4h
[14:02:45 CEST] <nonex86> but even if it still have frame basis encoding
[14:02:46 CEST] <hseg> BTW, any reason input streams don't support -to?
[14:02:59 CEST] <hseg> Hm.
[14:03:15 CEST] <nonex86> you can still cut from the first frame
[14:03:25 CEST] <nonex86> your audio stream can start earlier
[14:03:28 CEST] <nonex86> or later
[14:03:32 CEST] <nonex86> compared to video
[14:03:43 CEST] <nonex86> usually that gap is really small
[14:03:49 CEST] <nonex86> but in your case
[14:04:02 CEST] <nonex86> when you want delete from the middle of the file
[14:04:09 CEST] <nonex86> and then merge two parts
[14:04:14 CEST] <nonex86> i guess you need to reencode
[14:04:26 CEST] <nonex86> well, audio encoding is quite fast
[14:04:37 CEST] <nonex86> even if you need to resample stream
[14:05:31 CEST] <nonex86> even if you have several audio streams :)
[14:06:19 CEST] <nonex86> btw, by saying 'frame' here, i mean audio frame
[14:06:21 CEST] <hseg> Right, I noted that e.g. when youtube-dl uses ffmpeg to reencode audio or to merge multiple streams into one container, it's much faster than my current case of concating videos.
[14:07:36 CEST] <nonex86> in your case the good thing is you cut/merge the same video stream
[14:08:35 CEST] <nonex86> again, if you can tolerate idr/recovery point (key frame) accuracy
[14:08:46 CEST] <nonex86> you can just merge h264 NALU's
[14:09:04 CEST] <nonex86> if you need frame accuracy...
[14:09:18 CEST] <nonex86> that depends on video stream itself
[14:09:25 CEST] <hseg> OK, so I'm running ffprobe to get the keyframe positions.
[14:10:05 CEST] <hseg> So I guess my suggestion of blanking out that segment isn't useful?
[14:10:22 CEST] <nonex86> depends on you
[14:10:30 CEST] <nonex86> can you tolerate holes inside the stream? :)
[14:10:43 CEST] <nonex86> or you need to real cut/merge?
[14:10:56 CEST] <nonex86> cut 1st part using ffmpeg
[14:11:03 CEST] <nonex86> cut 2nd part using ffmpeg
[14:11:08 CEST] <nonex86> merge them using ffmpeg
[14:11:30 CEST] <nonex86> or maybe ffmpeg can do this thing in atomic operation
[14:11:32 CEST] <nonex86> dont know
[14:11:35 CEST] <hseg> I can tolerate image being black and sound being null at that point - it occurs right after a blackout.
[14:11:53 CEST] <nonex86> better ask cli adepts here :)
[14:11:56 CEST] <hseg> The only question is whether it will incur a long rendering time.
[14:14:02 CEST] <nonex86> https://trac.ffmpeg.org/wiki/Concatenate
[14:16:51 CEST] <hseg> I could cut and merge. However, doesn't that take a long time?
[14:17:07 CEST] <nonex86> depends
[14:17:19 CEST] <nonex86> dont forget about "copy" parameter
[14:18:31 CEST] <hseg> I've been using the concat filter... That may have caused my bad experience.
[14:19:29 CEST] <nonex86> :)
[14:22:42 CEST] <hseg> Yeah... So you're suggesting using -ss/-t to pick the parts before/after the split, passing -c copy to avoid reencoding, then concating via the concat protocol with -c copy?
[15:01:35 CEST] CTCP AVStream:  from beauty (beauty!~user at 2001:cc0:2026:4002:1423:827f:2a58:320a) to #ffmpeg
[15:01:35 CEST] <beauty> AVFormatContextAVCodecContext, which costs memory largely?
[15:06:39 CEST] <DHE> those data structures are not that big in the grand scheme of things. an active codec (decoder or encoder) may be huge though because the codec itself will store reference frames, etc.
[15:08:14 CEST] <beauty> oh.
[15:30:53 CEST] <fling> How do I add a covertart picture to ogg or mka properly? Should I set AV_DISPOSITION_ATTACHED_PIC somehow?
[15:42:43 CEST] Action: fling found vorbistools
[15:45:48 CEST] <Kirito> Many TV's now support interpolating (that's the correct term, right?) videos on the fly, making them appear as if they were recorded at higher frame rates by comparing two frames and generating a new one between them. This process is apparently inexpensive enough that it can be done in real-time, but it makes me wonder this:
[15:45:50 CEST] <Kirito> Could you gain much of a benefit by pre-encoding source videos with interpolation? For example, could you improve the quality of the interpolation much by comparing more than just two frames? Looking further ahead and farther behind when generating the middle frame
[15:46:42 CEST] <Kirito> and if so, are there any practical applications of this, specifically in ffmpeg filters (i.e. not proprietary)
[15:49:56 CEST] <DHE> TVs have DSP chips in them designed to do just that. it is fairly expensive in software, but you can build a chip to do it far quicker using custom circuitry
[15:50:13 CEST] <Kirito> Ahhh, interesting, that explains a lot
[15:50:27 CEST] <Kirito> As I was really surprised that TV's could do it so easily
[15:53:07 CEST] <Kirito> With that in mind still, could you still produce much better results in software? Or are DSP chips pretty much good enough that it wouldn't be worth the effort/extra storage
[15:57:44 CEST] <retard> this is the first thing to turn off
[15:57:51 CEST] <retard> when you get a new tv
[15:58:24 CEST] <Kirito> (Personal preference, I actually love it)
[15:59:36 CEST] <DHE> when playing video games you turn it off. no exceptions. giving the DSP frame look-ahead opportunities costs latency.
[16:00:04 CEST] <Kirito> Oh, that's fair definitely, I mean for regular TV though.
[16:46:21 CEST] <SpeakerToMeat> Good day.
[16:47:46 CEST] <SpeakerToMeat> Question, I'm mixing a video file and an audio file in a single mux (.mov) with ffmpeg using copy for both codecs (no recode), but ffmpeg and ffprobe are showing the wrong fps for the origin video file (showing 25 fps when I know for a fact the file is 24 fps), is there any way to change the fps (via constant frame count, that is reinterpeting) while muxing them?
[16:48:51 CEST] <SpeakerToMeat> I don't want a recode or processing of the frames, I just want ffmpeg to say "oh this is really 24fps not 25".
[16:50:31 CEST] <SpeakerToMeat> The input video file is DNxHD raw (extracted from an MXF with bmx)
[16:51:15 CEST] <nonex86> why are you care about fps?
[16:51:20 CEST] <nonex86> as long as pts is correct
[16:51:29 CEST] <nonex86> guess nothing to worrie about
[16:51:42 CEST] <nonex86> isnt it?
[16:53:18 CEST] <hidsbv> while joining two videos, I receive message- "Requested output format'concat' is not a suitable output format". whats wrong and how to correct it ?
[16:53:46 CEST] <c_14> concat isn't an output format
[16:53:49 CEST] <c_14> It's an input format
[16:54:31 CEST] <hidsbv> command - ffmpeg.exe -f concat -i lst.txt -c copy out.mp4
[16:55:05 CEST] <SpeakerToMeat> nonex86: pts?
[16:55:32 CEST] <nonex86> SpeakerToMeat: presentation timestamp
[16:55:50 CEST] <hidsbv> c_14: ?
[16:58:00 CEST] <hidsbv> while joining two videos, I receive message- "Requested output format'concat' is not a suitable output format". whats wrong and how to correct it ?
[16:58:02 CEST] <hidsbv> command - ffmpeg.exe -f concat -i lst.txt -c copy out.mp4
[16:59:18 CEST] <SpeakerToMeat> nonex86: Because I like my audio and video to be in sync
[17:03:51 CEST] <nonex86> SpeakerToMeat: your audio and video will be in sync if pts correct for both streams
[17:04:03 CEST] <nonex86> SpeakerToMeat: its not related to fps value you see in header :)
[17:04:11 CEST] <DHE> Kinda off-topic, I need to do some bitstream hacking to H264. Is there documentation or an ffmpeg module that can do SEI splitting and reassembly?
[17:04:35 CEST] <nonex86> SpeakerToMeat: ill ask again - why are you care about fps value in header? you have any problems with result file?
[17:14:00 CEST] <CorvusCorax> Hi. I want to extract individual frames from a video (as images) - but only the frames that are actually stored. For example, video has a timebase of 60 fps but only stores 20 fps, with possible gaps. I only want the frames that are actually there, if possible with the correct pts from the filename. is that possible?
[17:14:54 CEST] <CorvusCorax> I tried just recoding using the image2 codec, but that creates lots of extra images for all the "missing" frames, too
[17:22:53 CEST] <CorvusCorax> (which are all identical)
[17:35:00 CEST] <SpeakerToMeat> nonex86: There's no header, it's raw
[17:35:44 CEST] <SpeakerToMeat> nonex86: Yes, result file a) says it's 25 fps, we could load it on premiere and reinterpret it as 24 fps in a 24fps project there if it comes to that b) audio and video desync
[17:38:48 CEST] <nonex86> you have two separate streams
[17:38:52 CEST] <nonex86> audio and video
[17:39:01 CEST] <nonex86> and try to mux them in one file, correct?
[17:39:16 CEST] <SpeakerToMeat> correct
[17:39:48 CEST] <nonex86> i doubt your desync problem related to "wrong" fps you see in the output file
[17:40:05 CEST] <SpeakerToMeat> then what is it related to?
[17:40:17 CEST] <nonex86> start time in both streams are the same?
[17:40:25 CEST] <nonex86> or you provided time offsets somehow?
[17:41:01 CEST] <SpeakerToMeat> start time is the same, both files where synchronized inside the mxf when the video "knew" it was 24fps
[17:41:09 CEST] <SpeakerToMeat> and have the same count
[17:41:43 CEST] <SpeakerToMeat> DOn't worry I'll find this on my own, or fidn a tool that can do it. thanks anyhow
[17:41:53 CEST] <nonex86> ok :)
[17:44:49 CEST] <SpeakerToMeat> aaand the pts is wrong. it's at 25 fps
[17:51:36 CEST] <SpeakerToMeat> -r 24 on the input gives me a right pts, but it's not stored in the output file.
[17:54:06 CEST] <SpeakerToMeat> I'll just convert to discrete dpx frames and use that.
[18:03:10 CEST] <crunchybit> need to clean up some audio of a speech from 1968 of some hippy screaming in a megaphone. Sounds really compressed and distorted. Anything that ffmpeg and do?
[18:03:38 CEST] <crunchybit> if not any recommendations?
[18:20:41 CEST] <Amitari2> Anyone who can help me extract frames from a RAW AVI? (BGR)
[18:21:37 CEST] <Amitari2> This is what happens when I try.
[18:21:38 CEST] <Amitari2> http://pastie.org/10945514
[19:45:26 CEST] <CorvusCorax> @Amitari:  No decoder for stream #0:0, filtering impossible
[19:46:57 CEST] <CorvusCorax> that's odd, i had no issues with raw bgr24 data...
[19:47:04 CEST] <Amitari> CorvusCorax: Ouch.
[19:47:57 CEST] <CorvusCorax> and earlier: [avi @ 0x55719b365f20] Could not find codec parameters for stream 0 (Video: none (BGR[24] / 0x18524742), none, 2160x2160): unknown codec
[19:48:18 CEST] <JEEB> it just doesn't know what that format is supposed to be
[19:48:24 CEST] <CorvusCorax> maybe if you force the codec to raw, but I don't remember the syntax to enforce the decoding codec
[19:48:34 CEST] <JEEB> before -i you set -c:v rawvideo
[19:48:45 CEST] <JEEB> and then -pix_fmt BGR (also before -i)
[19:48:55 CEST] <JEEB> before -i = decoding parameters
[19:49:00 CEST] <JEEB> after -i = encoding parameters
[19:49:01 CEST] <CorvusCorax> thanks :)
[19:49:56 CEST] <JEEB> also you could just add that fourcc to work as raw video, BGR
[20:01:38 CEST] <Amitari> JEEB: I'll try!
[20:03:11 CEST] <Amitari> Nope, didn't work. :(
[20:03:15 CEST] <Amitari> http://pastie.org/10945577
[20:04:32 CEST] <JEEB> ouch
[20:06:01 CEST] <DHE> Does anyone know where to find code (or documentation, I can write my own if needed) on how to break an H264 stream down into its SEI components? I need to do a little bit of bitstream hacking
[20:07:25 CEST] <CorvusCorax> Amitari, do ffmpeg -pix_fmts
[20:07:54 CEST] <CorvusCorax> i think the one you need is "bgr24"   lower case
[20:08:00 CEST] <Amitari> Should that go before the -i, and what should it replace?
[20:08:06 CEST] <CorvusCorax> but that depends on how your raw video is encoded
[20:08:34 CEST] <CorvusCorax> its a separate call, with no other arguments, it just prints outr all supported pixel formats of your ffmpeg version :)
[20:08:46 CEST] <Amitari> Shit, something happened!
[20:09:27 CEST] <JEEB> well the error looks like pix_fmt isn't something that works with decoding (?) although that is still weird since you should be able to set pix_fmt with at least raw video input
[20:09:39 CEST] <Amitari> http://pastie.org/10945581
[20:09:58 CEST] <JEEB> yeah, it just parsed the pix_fmts option
[20:10:04 CEST] <JEEB> and printed you the available pix_fmts
[20:10:17 CEST] <CorvusCorax> thats what it was supposed to. those are the valid formats to be given with -pix_fmt
[20:10:22 CEST] <JEEB> yes
[20:10:31 CEST] <JEEB> he just gave it more parameters so he probably expected more than that from it
[20:10:33 CEST] <CorvusCorax> you need bgr24 I think
[20:10:51 CEST] <Amitari> So, how should the command go then?
[20:11:02 CEST] <Amitari> I sorta tried that myself, but I dunno what commands from before should be removed and whatnot.
[20:11:11 CEST] <JEEB> -pix_fmt bgr24 before -i
[20:11:13 CEST] <CorvusCorax> ffmpeg -c:v rawvideo -pix_fmt bgr24 -i video.avi /out/%08d.png
[20:11:43 CEST] <Amitari> Didn't work. :(
[20:11:52 CEST] <Amitari> http://pastie.org/10945583
[20:12:33 CEST] <JEEB> yeah, so it doesn't work for input weird enough :/
[20:12:57 CEST] <CorvusCorax> odd. but ffmpeg treats bgr24 as default for rawvideo avis anyway, just skip it
[20:14:33 CEST] <Amitari> You want me to try that?
[20:15:28 CEST] <CorvusCorax> ffmpeg -c:v rawvideo -i video.avi /out/%08d.png
[20:17:03 CEST] <Amitari> Well, something happened, but the conversion failed. :( http://pastie.org/10945586
[20:17:30 CEST] <Amitari> Wait!
[20:17:30 CEST] <Amitari> No!
[20:17:36 CEST] <Amitari> Now it works, I just had to change the filepath a litt.
[20:17:39 CEST] <Amitari> Thanks! :D
[20:19:33 CEST] <CorvusCorax> are the images good? or are they filled with pixel-puke?
[20:19:40 CEST] <BinaryBench> Hey, I'm on windows and do not seem to have a "screen-capture-device" as mentioned in https://trac.ffmpeg.org/wiki/Capture/Desktop
[20:19:48 CEST] <CorvusCorax> if the pix_fmt is wrong you can end up with weird results
[20:19:57 CEST] <Amitari> Nope, they look good! :D
[20:19:59 CEST] <furq> BinaryBench: https://github.com/rdp/screen-capture-recorder-to-video-windows-free
[20:20:18 CEST] <furq> that's the device referred to on the wiki, there are probably others you can use
[20:20:27 CEST] <furq> or use -f gdigrab which is slower but doesn't need anything external
[20:21:09 CEST] <JEEB> Amitari: that just happens to happen because the input pix_fmt is being defaulted to bgr24, so if it wasn't that it would probably be more problematic
[20:21:12 CEST] <BinaryBench> furq: speed is all I care about, however the phrase "java" gives me concerns about that....
[20:21:36 CEST] <furq> i've never used it but dshow should be much faster than gdi
[20:21:39 CEST] <JEEB> that said, there's a lot of semi-standardized fourccs for various raw things
[20:22:01 CEST] <BinaryBench> furq: I suppose I'll give it a go...
[20:22:09 CEST] <JEEB> (libavformat/riff.c has a whole long list of them)
[20:22:13 CEST] <CorvusCorax> While we are at converting videos to images, does anyone know a way to extract only existing frames from a video? Assume a camera stores an avi and claims 60fps but there's only 10 fps in it, if i do ffmpeg -i input.avi out%04d.png, I get lots of duplicate images where the missing frames get automatically duplicated
[20:22:16 CEST] <furq> it's probably worth trying gdigrab first
[20:22:19 CEST] <furq> it might be fast enough
[20:22:57 CEST] <CorvusCorax> ffmpegs -r  option to change framerate is only an option if the actually achieved rate is constant, but its not
[20:26:57 CEST] <BinaryBench> furq: with gdigrab, what would be the command to record?
[20:31:59 CEST] <bryan_> Hello everybody
[20:33:48 CEST] <bryan_> i am using ffmpeg interface API in my shared library...but when i run the third party application which uses my API...says undefined symbols for ffmpeg API...how can i resolve this issue?
[20:34:56 CEST] <JEEB> sounds like the library isn't in your search path?
[20:35:38 CEST] <bryan_> i have exported the ffmpeg library path in LD_LIBRARY_PATH on console but still it is not able to resolve it
[20:36:23 CEST] <JEEB> is it the same you linked against?
[20:36:27 CEST] <CorvusCorax> are you linking in your library dynamically or statically into the 3rd party app?
[20:36:28 CEST] <bryan_> i think one way to do is "dlopen" but does it really required?
[20:36:40 CEST] <JEEB> no, dlopen is not required
[20:36:40 CEST] <bryan_> dynamically
[20:37:09 CEST] <CorvusCorax> nah, dlopen you only need to load libraries fully dynamicaly, for exmaple to load plugins which aren't known at compiletime yet
[20:37:22 CEST] <JEEB> it just sounds like the version of the libraries you linked against aren't available
[20:37:55 CEST] <JEEB> I've been doing shared  builds in nonstandard locations for quite a while and WorksForMe is my experience so far
[20:38:19 CEST] <JEEB> so either it's something really dumbly simple or something very specific to what you're trying to do in some specific way
[20:38:20 CEST] <bryan_> ok
[20:39:32 CEST] <CorvusCorax> the systems dynmic linker should load the correct ffmpeg libs when your api library is being loaded. as long as they are installed
[20:40:29 CEST] <bryan_> hmm..will check it..might be doing something wrong in my setup..
[20:40:44 CEST] <bryan_> thank you
[20:54:39 CEST] <bencc> I'm doing several concurrent live transcoding tasks. what cpu metric tells me I doing to much and might drop frames?
[20:56:37 CEST] <BtbN> none, the stream fps do.
[20:57:21 CEST] <bencc> BtbN: ok
[20:58:14 CEST] <BtbN> you can probably guess that when all cores are at 100%, you are likely also dropping frames. But it might be just allright then as well. Or it might be dropping earlier, because it's not perfectly distributed accorss all cores.
[21:53:03 CEST] <DHE> bencc: that's something you should measure ahead of time and test run. BtbN pretty much nails it - only an actual measurement from ffmpeg will provide the information.
[21:53:57 CEST] <BtbN> Can probably also take a look at the outgoing bandwidth
[21:54:10 CEST] <BtbN> if you are doing strict CBR, you are able to calculate the expected bandwidth quite accurately
[21:54:15 CEST] <DHE> I find that at 95% usage you might run into a CPU cap. operating systems don't migrate threads real-time so you can end up with 0 active jobs on one core and 2 on another. you're effectively bottlenecked
[22:16:44 CEST] <teratorn> is anyone very familiar with RTSP over UDP? how does the stream / player / receiver recover from udp packet loss, and how does that translate in to loss of video and audio frames? is there any retransmit logic in the protocols? we're transcoding from RTSP to HLS and the HLS files are  getting corrupted due to lost packets in the incoming RTSP/UDP stream...
[22:20:28 CEST] <BtbN> It will be fixed on the next I frame.
[22:22:50 CEST] <DHE> I'd suggest checking the packet spacing on either the receiver or (preferably) the sender side. make sure the packets are spaced and not in sudden bursts
[22:23:38 CEST] <DHE> I've had a lot of problems with that all across the network.
[22:26:28 CEST] <teratorn> DHE: well I don't think I have any control over the sender, it's a little boxy encoder thing
[22:26:43 CEST] <teratorn> BtbN: could you elaborate?
[22:27:09 CEST] <BtbN> that's how encoding works. If a frame is missed/broken, the stream is broken until the next I frame comes by
[22:33:22 CEST] <DHE> key frames (I frames) are full frames, other frames are based on the differences to other frames
[22:41:10 CEST] <SchrodingersScat> is there a format that's better than others for getting clean cuts?  like not having it go to a keyframe or etc?
[22:41:38 CEST] <DHE> well, if you're willing to crank the bitrate you could make every frame a keyframe...
[22:41:46 CEST] <SchrodingersScat> tell me more
[22:42:11 CEST] <DHE> with ffmpeg using -g 1 uses nothing but keyframes
[22:42:17 CEST] <SchrodingersScat> is very short, so size isn't really an issue
[22:43:11 CEST] <teratorn> BtbN: OK I guess that makes sense
[22:44:11 CEST] <SchrodingersScat> ahahahaha, I may have spoken too early, this is gonna be huge, but should be fine.  DHE thanks for the tip.
[22:44:30 CEST] <DHE> SchrodingersScat: you were warned. :)
[22:44:49 CEST] <SchrodingersScat> quite
[22:45:43 CEST] <DHE> SchrodingersScat: might compromise on a small number, like -g 5 depending how accurate you need
[22:46:01 CEST] <furq> you could just use an intra-only codec
[22:46:13 CEST] <DHE> like mjpeg?
[22:46:16 CEST] <furq> that's probably better if you're going to be cutting it up later
[22:46:23 CEST] <furq> i was thinking more like ffv1 but sure
[22:46:28 CEST] <DHE> after teratorn's issue I had keyframes on the brain...
[22:46:52 CEST] <DHE> ffv1 is all keyframes? good to know...
[22:47:01 CEST] <furq> it is by default
[22:47:09 CEST] <furq> you can increase the gop size but i don't know how that actually works
[22:47:22 CEST] <furq> it doesn't seem to reduce the bitrate very much
[22:49:42 CEST] <SchrodingersScat> k, think that took care of it
[22:56:27 CEST] <teratorn> BtbN: so what is the trick to not corrupting the HLS stream that I'm transcoding to? I guess you MUST NOT encoding any frames until you get another  valid iframe again?
[22:56:58 CEST] <BtbN> Don't lose any packets I guess
[22:56:58 CEST] <DHE> if you're getting packet loss, it means a network problem. your hardware should be able to tell you where it's happening
[23:03:56 CEST] <teratorn> U D P
[23:04:04 CEST] <teratorn> rtsp supports streaming over U D P
[23:04:13 CEST] <teratorn> you will get packet loss
[23:04:26 CEST] <teratorn> I'm not sure why I *have* to use UDP instead of TCP, but there yo ugo
[23:04:43 CEST] <teratorn> I think the packet loss (wifi) was causing the tcp streams to hang, so we switched to udp as a quick fix (i'm sorry)
[23:04:49 CEST] <teratorn> for making the internet worse
[23:04:52 CEST] <BtbN> could try to use a container with solid error correction. But I don't think ffmpeg supports muxing that.
[23:05:08 CEST] <BtbN> UDP is the better choice for streams usually
[23:05:17 CEST] <teratorn> i don't have any choice, it's either rtsp over tcp or rtsp over udp from this boxy encoder gadget
[23:05:19 CEST] <BtbN> With a short gop interval, you loose a few seconds at max
[23:05:41 CEST] <teratorn> i'm not sure how configurable the boxy thing is, or even if we are allowed to configure it. i doubt it.
[23:06:19 CEST] <DHE> ... wifi?
[23:06:29 CEST] <teratorn> it's a matter of disregarding all the necessary data from the corruption caused to the rtsp stream due to packet loss, and encoding a valid hls stream just with placeholder frames
[23:07:22 CEST] <teratorn> DHE: yes, hanging tcp rtsp streams due to wifi packet loss. now having switched to udp, our transcoding is borked because we're not actually handling hte packet loss correctly for some reasons.
[23:07:23 CEST] <DHE> it's not quite that easy. from the first corrupted frame, some kind of image smear will persist until the next keyframe
[23:07:41 CEST] <teratorn> nothing I can do about that rtp doesn't have any retransmit logic built in
[23:07:54 CEST] <teratorn> i don't control the sender. black box rtsp stream over tcp or udp...
[23:08:06 CEST] <teratorn> so you either use tcp
[23:08:11 CEST] <DHE> can you make it more reliable? plug something into the same switch as this encoder thing?
[23:08:13 CEST] <TD-Linux> rtcp supports retransmissions. you should be able to negotiate that with rtsp
[23:08:15 CEST] <teratorn> or you use udp and recover from packet loss *somehow*
[23:08:18 CEST] <furq> is the sender using wifi
[23:08:38 CEST] <teratorn> TD-Linux: thank you I will have to look in to rtcp to see if that is supported
[23:09:12 CEST] <teratorn> and how we can use it to recover lost frames... but keep in mind this is live streaming, so the device would have to maintain a buffer to retransmit frames from
[23:09:14 CEST] <BtbN> it's not trivial, if possible at all, to detect that kind of corruption
[23:09:19 CEST] <teratorn> s/frames/packets/
[23:09:49 CEST] <TD-Linux> BtbN, yes it is, RTP has sequence numbers
[23:09:59 CEST] <teratorn> yes, the sequence numbers, but!
[23:09:59 CEST] <DHE> I'm suggesting you prevent the loss rather than recover from it. why are you using wifi here?
[23:10:04 CEST] <teratorn> udp may be delivered out of order
[23:10:07 CEST] <BtbN> That's if a whole frame gets lost?
[23:10:11 CEST] <TD-Linux> teratorn, there are a lot of ways to do it, and yeah you need to keep them around for retransmission
[23:10:14 CEST] <teratorn> DHE: no choice
[23:10:26 CEST] <TD-Linux> there are other things you can do with RTCP, like request an Iframe on packet loss rather than a retransmission
[23:11:00 CEST] <teratorn> DHE: I can't control all the variables. I just have to make a solution that lets us "recover" from udp packet loss of the rtsp stream, either by inducing retransmission via an rtcp control channel, or by disregarding data and only doing the transcode with valid data
[23:11:15 CEST] <teratorn> TD-Linux: OK i definiately have homework to do, thanks
[23:11:40 CEST] <teratorn> TD-Linux: its a little boxy encoder thing, I'm not sure how much control I have
[23:11:47 CEST] <TD-Linux> there is even really fancy stuff like request a frame predicted only from frames you have, though your options might be limited as you don't have much control over one side
[23:12:12 CEST] <teratorn> basically, if I can do it with well supported rtsp/rtcp features, I'm ok
[23:13:02 CEST] <teratorn> or, again, just ignore the frames that are corrupted or have references to missing frames, and once enough time has elapsed, I can go ahead and pretend those packets will never come in, mark them as lost, and continuing the transcode with placeholder frames
[23:16:09 CEST] <TD-Linux> teratorn, yup, that will work too (waiting for next keyframe). if it helps to google, "AVPF" is what you need in the rtsp sdp
[23:16:18 CEST] <TD-Linux> (for rtcp)
[23:20:40 CEST] <teratorn> TD-Linux: thank you
[23:40:44 CEST] <SchrodingersScat> yay, i figured it out, thanks guys
[23:41:40 CEST] <DHE> Does anyone know where to find code (or documentation, I can write my own if needed) on how to break an H264 stream down into its SEI components? I need to do a little bit of bitstream hacking.
[23:51:55 CEST] <emilsp> what am I doing wrong if when I parse my avfilter (I need to crop a frame), avfilter_graph_parse2() returns a nonzero code and spouts in stderr [AVFilterGraph @ 0x7ff801e2c320] No such filter: 'buffer'
[23:52:47 CEST] <emilsp> http://sprunge.us/ZJBb?c on line 48
[23:53:56 CEST] <DHE> you don't need to provide buffer and buffersink in the pipeline. you create them yourself as endpoints to the pipeline and attach them manually
[23:53:58 CEST] <jpsharp> I'm feeding ffmpeg with -i file.gif with an output to an mp4 file.  I have it set for infinite loop.  That all works, but I'd like to be able to just change the gif while the process is running and have the resultant mp4 file change.  ffmpeg seems to read the file once and then use that over and over, how can I tell ffmpeg to reload the file either on command or occasionally.
[23:54:35 CEST] <emilsp> DHE, so the 'arg' string is wrong ?
[23:54:59 CEST] <DHE> emilsp: there's some examples in the doc directory, including an avfilter
[23:55:09 CEST] <emilsp> woohoo, thanks :)
[23:55:26 CEST] <DHE> doc/examples/filter*.c
[23:56:30 CEST] <emilsp> yup, am reading them already, thank you very much
[23:58:47 CEST] <t4nk834> hi all
[00:00:00 CEST] --- Wed Oct 19 2016