[Ffmpeg-devel-irc] ffmpeg.log.20130610

[burek]

[00:00] <AKX> Ah well, I'll see what it did in the morning :D
[00:14] <AKX> Hmmh, no, the second video in the sequence gets audio/video desync. Ah well, have to figure out why tomorrow
[05:02] <vstemen> Hi.  Since upgrading ffmpeg from 0.6.1 to 0.7.15, in the process of upgrading FreeBSD, I now get the error "File for preset 'veryfast' not found" with the "-vpre veryfast" option.
[05:03] <vstemen> I notice that preset file no longer exists in the package under /usr/local/share/ffmpeg
[05:05] <vstemen> I notice a comment in the change log that says "FFmpeg now accesses x264 presets via libx264" and the help output for the x264 command shows 'veryfast' as a --preset option, yet ffmpeg doesn't work with it.
[05:37] <vstemen> Problem solved.  Apparently you have to use '-pre' now rather than '-vpre' in order to pass the option to x264.
[07:52] <SirCmpwn> can I use ffmpeg to convert soft subs into hard subs?
[07:59] <ubitux> you can use -vf subtitles filter
[07:59] <ubitux> or -vf ass
[08:00] <SirCmpwn> thanks, I'll use that as a guide to search through these massive man pages
[08:01] <SirCmpwn> ubitux: I don't see "subtitles" or "ass" in the list of video filters in the man pages
[08:06] <alin|mobile> hey there; i am currently using an asus eeepc, apu e-450 and integrated radeon hd 6k; i am still struggling with the frequencies, because my graficC. only runs with 275 Mhz instead of 600; any ideas how to get it work using the full scale? btw. using the latest catalyst suite; thx in advance
[08:07] <DeXa> SirCmpwn: http://www.ffmpeg.org/ffmpeg-filters.html#ass
[08:08] <DeXa> SirCmpwn: make sure you enabled --enable-libass in your build
[08:08] <SirCmpwn> thank you, DeXa
[08:08] <SirCmpwn> what if my subtitles are embedded in an MKV file
[08:08] <SirCmpwn> looks like both the ass and subtitles filter require a second file
[08:09] <ubitux> you have to extract them first, or use just pass the video as parameter
[08:09] <SirCmpwn> ah, fancy
[08:09] <SirCmpwn> thanks for the help
[08:09] <ubitux> that's a current limitation
[08:09] <ubitux> until subtitles are inserted in libavfilter properly
[08:09] <DeXa> ffmpeg -i file.mkv and look for the subs stream [x]
[08:10] <SirCmpwn> I'm finding myself building more and more software from source
[08:10] <SirCmpwn> bloody repository maintainers
[08:10] <DeXa> I feel you. homebrew save me from that :)
[10:31] <Olson> on the off chance I'm wondering what might be cause of window flickering when recording a screencast with ffmpeg + X11grab - the window i'm recording has occasional sort of scanlines on it, revealing the desktop beneath, anyone see something like that?
[10:34] <Olson> I dunno if that's clear http://getblended.org/screens/videoglitch.jpg the window is recorded and the glitch at the top reveal the desktop "underneath"
[11:18] <jennifer> Hi, once I install ffmpeg, what do I do to remove the sound / audio from a mp4 file that I have?
[11:20] <Mavrik> ffmpeg -i <input> -an -codec:v copy output.mp4
[11:25] <jennifer> That's it huh..
[11:28] <Mavrik> yep, "codec:v" tells it to just copy video, -an means "disable audio"
[11:29] <jennifer> Can it be enabled later on?
[11:29] <jennifer> I hope not..
[11:31] <Mavrik> uh, maybe I should say "disable audio copy"
[11:31] <Mavrik> :)
[11:43] <jennifer> :) Okay thank you.
[11:46] <Guest53085> Hi again, last Friday I make a question as Guest35974 about the releasing time of a specific version of Lavf
[11:46] <Guest53085> Just came back to thank the members of this channel, specially llogan, saste and fflogger for his guidance that allowed me to find the solution using git log and git show.
[11:46] <Guest53085> Sept-2012. Problem solved, many thanks :)
[11:52] <gangam> anyone knows how to stream multicast rtp and open in vlc ?
[12:25] <braincracker> hi
[12:25] <braincracker> https://en.wikipedia.org/wiki/Binary_Golay_code  why is it cool to shuffle the redundancy matrix?
[14:34] <fintikliuszka> Hello, i have a question regarding development using libav. I want ffmpeg to decode video frame into OpenGL pixel buffer object, without using memcpy function. Is it possible to give ffmpeg a pointer so that it would fill that memory with video frame data?
[15:40] <iive> fintikliuszka: you can, but you should research get_buffer/release_buffer functionality. It is called direct rendering. Another problem is that most codecs (all mpeg1/2/4) use YUV colorspace, while OpenGL mostly expects RGB textures.
[15:42] <iive> you can use swscale for yuv->rgb . Another possibility is to do the conversion in the video card, e.g. mplayer -vo gl  supports blend trick or a number of pixel shader conversions.
[15:42] <fintikliuszka> iive: thank you, i think i can do it in shaders.
[15:43] <fintikliuszka> what about the AVFrame structure, why does it contain array of data?
[15:44] <fintikliuszka> i cannot get it
[15:44] <fintikliuszka> why not just one pointer?
[15:44] <iive> yuv is usually processed each channel seprately.
[15:44] <iive> it is like having one gray bitmap for Y, one for U and one for V
[15:45] <iive> for YUV420, U and V channels are half vertical and half horrizontal resolutions.
[15:45] <Mavrik> most new GPUs have hardware accelerated YUV conversions though
[15:45] <iive> so these "bitmaps" are called planes.
[15:46] <iive> Mavrik: they all do, but I'm not aware of opengl primitive for it. and even it could be "shader" implementation.
[15:46] <Mavrik> of course. :)
[15:47] <Mavrik> all new OpenGL stuff is done via shaders :)
[15:47] <Mavrik> I know Android's OGL ES implementation has an accelerated texture type which does conversion automatically
[15:47] <Mavrik> so worth checking into that to avoid doing conversions on CPU
[15:48] <fintikliuszka> now i get it :) thanks
[17:39] <jameshowe> I'm having some problems with the new tee muxer
[17:39] <jameshowe> http://pastebin.com/jbGPrEaQ
[17:39] <jameshowe> Tried to convert my three previous commands to a single one, but it's not happy
[17:42] <jameshowe> oh, and I actually used execvp, so imagine quotes around all those arguments
[17:46] <jameshowe> aha, I was missing a -map
[17:49] <jameshowe> so now the question becomes, how do I add the aac_adtstoasc filter?
[17:50] <jameshowe> because putting it in the format options doesnt work
[17:58] <jameshowe> or equivelently, h264_mp4toannexb
[18:00] <jdolan> is there a way to force each input file to a specific framerate?
[18:00] <jdolan> i have a .flv file with a variable framerate, in fact the video stream in the .flv file stops for some time while the audio keeps going.
[18:00] <jdolan> this confuses the crap out of ffmpeg.
[18:01] <jdolan> mplayer can play the .flv back just fine.
[18:05] <jdolan> if i play back the .flv with ffplay, it never stops. it thinks the audio stream keeps going to infinity.
[18:05] <jdolan> this does not happen with mplayer.
[18:05] <jdolan> mplayer correctly plays the .flv to the end of both streams and exits cleanly.
[18:06] <jdolan> i can upload the .flv file if someone cares to check.
[18:06] <jameshowe> If I set global_header it asks for h264_mp4toannexb, if not it askes for aac_adtstoasc
[18:06] <jameshowe> Is it actually possible to tee to mp4 and mpegts in one command?
[18:34] <jdolan> weird. i'm running ffmpeg-devel from macports, which is quite recent.
[18:34] <jdolan> yet -vsync drop is apparently not supported.
[18:34] <jdolan> was it recently removed?
[18:34] <jdolan> it was just added last year..
[20:47] <jdolan> ubitux: i have a sample .FLV that reproduces my problem.
[20:48] <jdolan> mplayer can play this back fine, ffplay plays it but then doesn't realize it should stop and just plays silence ad infinium.
[20:48] <ubitux> -auto-exit
[20:48] <ubitux> -autoexit sorry
[20:49] <jdolan> http://jdolan.dyndns.org/tmp/jay_and_gonzo.flv
[20:49] <jdolan> okay i'll try that.
[20:49] <jdolan> okay that works.
[20:50] <jdolan> is that not the default behavior or is that just helping this video behave correctly?
[20:50] <jdolan> seems odd that i should have to tell it that (?)
[20:51] <jdolan> seeing ffplay replay that video correctly is encouraging.
[20:51] <ubitux> ffplay doesn't exit automatically
[20:51] <ubitux> with every media
[20:52] <jdolan> my real issue is that i have two flv's, each of which might have missing video frames (like the one i linked to), and i'm trying to overlay these videos side by side.
[20:52] <jdolan> what i'd like to do is always sync each input video to its audio track.
[20:52] <ubitux> if you want help on the -vsync drop issue, you have to pastebin like asked btw
[20:52] <jdolan> in other words, duplicate video frames to keep in sync with audio when the video drops out.
[20:52] <jdolan> clearly ffplay can do this, as i can watch the video linked above now just fine.
[20:53] <jdolan> but ffmpeg gets confused when filtering 2 such videos together.
[20:53] <jdolan> sure i'll pastebin my script, one second! :)
[20:53] <ubitux> "Please don't paste your scripts"
[20:54] <jdolan> http://pastebin.com/sKGQuy5b
[20:55] <jdolan> oh, sorry. the script is literally just an ffmpeg call.
[20:55] <ubitux> we need "the COMPLETE console output"; also, samples would help
[20:55] <jdolan> sure, gladly. one second.
[20:56] <ubitux> i think i've seen that command a while ago on ffmpeg-user
[21:02] <jdolan> wasn't from me.
[21:03] <jdolan> here's the whole thing, complete with examples ready to go (http)
[21:03] <jdolan> http://pastebin.com/TrsnuScu
[21:03] <jdolan> you can remove -vsync drop and the command works, and produces some sort of merged .mp4.
[21:03] <jdolan> but you can see the timesync is hosed.
[21:04] <jdolan> jay_and_gonzo.flv is missing video frames.
[21:04] <jdolan> jay.flv is not.
[21:04] <jdolan> but in my scenario, each video will always have complete audio streams, and each video should sync against the video timecode only.
[21:05] <jdolan> and duplicate video frames as necessary. that's what i'm hoping to achieve, anyway.
[21:06] <jdolan> if you drop the -vsync option, you'll want to add -r 15 or something
[21:10] <ubitux> indeed there is a nice time wrap created by that pic
[21:10] <ubitux> n:102 pts:5145 pts_time:5.145 pos:325373 fmt:yuv420p sar:0/1 s:320x240 i:P iskey:1 type:I checksum:41AC392B plane_checksum:[C2155D55 FD2127DD 11A9B3EA]
[21:10] <ubitux> n:103 pts:12661 pts_time:12.661 pos:386598 fmt:yuv420p sar:0/1 s:320x240 i:P iskey:0 type:P checksum:7BB2464B plane_checksum:[EED1DBAE 554CE81E F1FB8261]
[21:11] <ubitux> the picture should be repeated IMO
[21:11] <ubitux> i don't know you generated it but i would recommend to change that if you can
[21:15] <jdolan> hm.
[21:15] <jdolan> i'm not sure that i can.
[21:15] <jdolan> if you play the video back by itself, it plays fine.
[21:15] <jdolan> in either ffplay (-autoexit) or mplayer.
[21:15] <jdolan> it also streams fine into the flash client.
[21:15] <ubitux> yes but as you can guess there are some synchronization issues to deal with when overlaying
[21:15] <jdolan> the only thing that fails is ffmpeg merging it.
[21:16] <jdolan> right, but is there a way to let audio TC drive all syncing?
[21:16] <ubitux> maybe something is possible with the asyncts filter
[21:17] <jdolan> i'll look
[21:17] <ubitux> i don't know if there is a pretty way to do that currently anyway
[21:17] <ubitux> feel free to open a trac issue
[21:18] <jdolan> just curious, any idea why the -vsync drop isn't available to me?
[21:18] <ubitux> i must say i don't really know how vsync works
[21:20] <jdolan> no worries then, thank you for your help.
[21:29] <jdolan> ubitux: oh yea, this other thing caught my eye: http://www.ffmpeg.org/ffmpeg-all.html#Advanced-options
[21:30] <jdolan> -map apparently takes a sync file -- awesome! i can tell ffmpeg to sync my [out0] against [0:1] right?
[21:30] <jdolan> false. apparently the [linklabel] syntax doesn't support the syncfile. LOL.
[21:31] <jdolan> it's like there's some blocker down every path i try.
[21:34] <ganagm> Hi, is it possible to configure ffmpeg with x264 to use maxbitrate and VFR so it will reduce frame rate when maxbitrate is reached ?
[21:35] <JEEB> no, at least not easily or without hacking
[21:35] <JEEB> x264 is a very capable encoder and it is made to keep your rate no matter what, and not drop a single frame you input into it
[21:36] <JEEB> so all the dropping should be done within libavcodec and friends, yet it should have to be able to see the result of the encode or whatever :P
[21:36] <JEEB> if you want VFR, you should provide a VFR source and that'd be the simplest way
[21:37] <JEEB> if you're doing video over a limited bandwidth I hope you know about maxrate and bufsize settings, those are a must :P
[21:46] <ganagm> JEEB: I am doing live video produced from screen capture and sent to many clients, that takes alot of bandwidth, I am using maxbitrate but unfortunatly it lowers the quality very much. I was hoping to find a way to lower frame rate when needed
[21:47] <JEEB> try heightening bufsize? That is, if you can adjust it on the clients' side as well
[21:47] <JEEB> otherwise slower settings might be useful, if possible
[21:49] <ganagm> what do you meen by slower settings ? preset and tune ?
[21:50] <JEEB> -preset
[21:53] <ganagm> OK, thanks for the tips. btw currently I am using vlc to serve as an rtsp server. is it possible to do it from ffmpeg itself. I want to reduce the overhead of sending to vlc ( udp ) ?
[22:03] <jdolan> ubitux: i think what i'll do is let ffmpeg fix each video first.
[22:03] <jdolan> ubitux: it's extremely quick to just pipe the vids through ffmpeg and let it rewrite the TC
[23:29] <Olson> mm I was using libfaac as audio codec for my things, is that not recommened any more? it says I dont have it
[23:29] <Olson> Ihave now libfdk_aac
[23:33] <JEEBsv> Olson: fdk-aac is much better than faac
[23:34] <hack_sesh> If I use the overlay filter to merge two clips and put them side by side, is there any way to say 'dont start playing the second clip until a delay of 3 seconds'?  I can use the 'enable' param to not have it appear for 3 seconds, but when it does its starting at t=3, not t=0.
[23:34] <Olson> JEEBsv, cheers, so this is ok then what I use there.. it said "experimental" or something as things were whizzing by, perhaps I am mistaken
[23:34] <yajiv> Hi i'm trying to run a low latency encode and i'm seeing something thats' a little bizzare. I'm noticing that time between capture from fbdev up until encoding is around 20-30ms however the rest of the process, including the encode, through to tcp write out is around 5ms.
[23:35] <JEEBsv> Olson: the libavcodec internal aac encoder has that flag
[23:35] <JEEBsv> as it's still not fully developed to be effective
[23:35] <yajiv> i'm really surprised to see the long time taken from capture up until encoding begins and i was curious if someone had some pointers.
[23:35] <yajiv> it's on version : ffmpeg version N-52509-g785eb5f
[23:35] <JEEBsv> fdk-aac doesn't have the experimental flag
[23:36] <Olson> ffmpeg -codes | grep aac  is showing libfdk_aac  and aac_latm
[23:37] <Olson> maybe when I build.. I didn't enable something, I just followed some instruction
[23:37] <JEEBsv> if you have fdk you have fdk
[23:37] <JEEBsv> and that is currently the best thing around
[23:37] <JEEBsv> (fraunhofer's encoder that was open source'd by google)
[23:38] <JEEBsv> too bad it's not compatible with (L)GPL so you can't redistribute the binaries :<
[23:38] <Olson> oh I see ok, cool I'll just use that then cheers
[23:38] <JEEBsv> just make sure you have afterburner on :)
[23:39] <JEEBsv> it's some random option that makes it crunch a bit more when compressing the audio for maximum effect
[23:39] <Olson> I dont suppose.. is that that -ab flag
[23:39] <Olson> oh no.. audiobitrate
[23:39] <JEEBsv> -b:a is the new way of setting the bit rate btw :)
[23:40] <JEEBsv> so you have the option, and then you can stick :v or :a etc. to it to point towards something specifically
[23:40] <JEEBsv> afterburner should be just -afterburner 1
[23:40] <JEEBsv> when you have fdk-aac built in
[23:41] <Olson> so -b:a 128k  for the audio (in this case)
[23:41] <Olson> it's just voice, like
[23:41] <JEEBsv> yup
[23:41] <Olson> noice
[23:42] <Olson> must excuse me.. I been using same copy and pasted commandline for donkeys because it just kept working (more or less) :P
[23:42] <JEEBsv> and *codec can now be replaced with just -c:v or :a or :s f.ex.
[23:43] <Olson> ohmm..ow ok.
[23:43] <Olson> very modern :)
[23:44] <Olson> I will try this. thanks
[23:46] <yajiv> sry..didn't mean to jump in the middle of that conversation so i'll re-add the question:
[23:46] <yajiv> Hi i'm trying to run a low latency encode and i'm seeing something thats' a little bizzare. I'm noticing that time between capture from fbdev up until encoding is around 20-30ms however the rest of the process, including the encode, through to tcp write out is around 5ms.
[23:46] <yajiv> i'm really surprised to see the long time taken from capture up until encoding begins and i was curious if someone had some pointers.
[23:46] <yajiv> it's on version : ffmpeg version N-52509-g785eb5f running on android
[23:59] <yajiv> and sometimes even 100ms
[00:00] --- Tue Jun 11 2013