[Ffmpeg-devel-irc] ffmpeg-devel.log.20130419

burek burek021 at gmail.com
Sat Apr 20 02:05:02 CEST 2013


[00:00] <ubitux> saste: the latest version is available on my github (filter-timeline branch)
[00:00] <saste> ubitux, not blocking
[00:00] <ubitux> ok
[00:00] <saste> but lemme have another look
[00:00] <ubitux> you want the patchset on the mailing list or the remote is fine?
[00:00] <saste> (and give nicolas a chance to reply)
[00:01] <saste> ML is better
[00:01] <ubitux> ok
[00:02] <ubitux> durandal_1707: didn't you commit a filter with a frame counter recently?
[00:02] <ubitux> or it's still a pending patch?
[00:03] <ubitux> mmh that branch doesn't look updated
[00:04] <ubitux> saste: sorry i'll push a fixed version in a moment
[00:06] <ubitux> saste: here you go
[00:06] <ubitux> should be better.
[00:06] <ubitux> only 3 patches now
[00:07] <ubitux> i'll send to the ml them in a moment
[00:46] <saste> ubitux, will review tomorrow before/during gsoc time, now i need to sleep
[00:47] <ubitux> good night :)
[00:48] <saste> review my interleave if you're bored ;-)
[00:51] <ubitux> i'm not, i'm still playing with gif
[00:51] <ubitux> but i'll do it
[00:52] <ubitux> i've started my 6th branch for gif
[00:52] <ubitux> would be nice to have some review for the patches i sent, notably fate :(
[00:52] <ubitux> (and timestamp fix!)
[00:58] <durandal_1707> ubitux: isn't that frame counter thing same just in another place, that does not improve situation when pts does not start from 0
[00:59] <ubitux> yes
[00:59] <ubitux> the problem is just in a single place noew
[00:59] <ubitux> -e
[00:59] <ubitux> but the patch was motivated by the frame number export mainly
[00:59] <durandal_1707> also why adding .passthrough_filter_frame cant it be done with .filter_frame ?
[01:00] <ubitux> how?
[01:02] <durandal_1707> nvm, not that important
[01:02] <durandal_1707> i would really like to see multi-thread support
[01:03] <ubitux> in libavfilter?
[01:03] <ubitux> or in the gif encoder?
[01:03] <ubitux> :D
[01:04] <durandal_1707> yes
[01:04] <durandal_1707> git encoder - no ...
[01:05] <ubitux> :D
[01:05] <ubitux> it will be pretty hard anyway :p
[01:05] <durandal_1707> both slice and frame, like in lavc
[01:05] <durandal_1707> why it will be hard?
[01:05] <ubitux> for gif?
[01:06] <durandal_1707> omg, forget git
[01:06] <ubitux> because of previous frames i guess
[01:06] <durandal_1707> *gif
[01:06] <ubitux> :)
[01:06] <ubitux> lavfi doesn't support slice threading anymore
[01:06] <ubitux> frame threading for a start would be nice
[01:06] <durandal_1707> slice threading is just calling mt inside filter ....
[01:07] <ubitux> actually, any filter flagged with the timeline flag without custom callback can be threaded
[01:07] <durandal_1707> "slice" is just name....
[01:07] <ubitux> so as soon as i pushed the timeline thing
[01:07] <ubitux> you can consider threading in all of them
[01:07] <ubitux> relatively trivially i guess
[01:08] <ubitux> but it would be nice to cleanup the get_buffer thing first
[01:09] <durandal_1707> what needs cleanup?
[01:10] <durandal_1707> i think it would be better if i do not modify interlacet/tff in joinfields and leave that to another filter
[01:10] <ubitux> filters setting callbacks unecessarily
[01:10] <ubitux> or filters not having it where required eventually
[01:10] <ubitux> and... frame queue :)
[01:11] <ubitux> pool*
[01:13] <ubitux> http://pastie.org/7651205
[01:13] <ubitux> :)
[01:14] <ubitux> maybe i should start disposals for good now
[01:14] <ubitux> durandal_1707: care to review a few gif patches?
[01:14] <ubitux> notably fate which is blocking, and eventually gif timestamps
[01:14] <ubitux> ...and if you need some review... :)
[01:15] <ubitux> ah and i need the palette patch to start disposals
[01:15] <ubitux> :p
[01:16] <durandal_1707> ubitux: how much gce patch saves?
[01:16] <ubitux> not much i believe, lemme check
[01:16] <ubitux> it only saves some space with pal8 btw.
[01:17] <ubitux> but it simplifies the code, and it helps me :p
[01:17] <cone-523> ffmpeg.git 03Nicolas George 07master:b42827ffd99f: lavfi/vf_aspect: improve compatibility of parsing.
[01:17] <cone-523> ffmpeg.git 03Nicolas George 07master:b1cc12d0e3aa: ffmpeg: make -aspect work with -vcodec copy.
[01:17] <cone-523> ffmpeg.git 03Michael Niedermayer 07master:7254afdf1c60: Merge remote-tracking branch 'cigaes/master'
[01:18] <ubitux> oh cool, new patch from the vidstab guy
[01:27] <cone-523> ffmpeg.git 03Reimar Döffinger 07master:a39cd8766fba: Fix multithreaded MPEG-4 decoding.
[01:31] <cone-523> ffmpeg.git 03Marton Balint 07master:007989c7a221: mxfdec: fix return variable type for avio_seek
[01:37] <durandal_1707> if would be nice if ffmpeg -h filter=lut outputs supported formats for inputs and outputs
[01:37] <ubitux> yeah
[01:37] <ubitux> but..
[01:37] <ubitux> :)
[01:38] <ubitux> not easy i'd say
[01:38] <durandal_1707> ah, because its dynamic thing even for cases it should not be....
[01:38] <ubitux> yes
[01:53] <durandal_1707> i should write planeextract/extractplane
[01:56] <Compn> is there an easy way to put timestamps on videos yet ?
[01:56] <Compn> like what in cctv footage
[01:57] <Compn> that might be a useful filter
[01:57] <Compn> if not
[01:57] <iive> timecode
[01:58] <durandal_1707> with drawtext, i couldn't get it to work, i only get pts
[02:00] <ubitux> %{pts}
[02:00] <ubitux> but a timed representation instead would be nice
[02:00] <durandal_1707> yes, pts works, but date/time from doc doesn't
[02:00] <ubitux> durandal_1707: it does, somehow
[02:01] <ubitux> (use basetime)
[02:01] <ubitux> i've sent a patch a long while ago
[02:01] <ubitux> but it was likely going to break on some systems
[02:03] <ubitux> thx for the reviews durandal_1707 
[02:05] <ubitux> don't forget the timing one! :)
[02:18] <cone-523> ffmpeg.git 03Clément BSsch 07master:3d786591b814: fate/gif: create meaningful gif encoding tests.
[02:18] <cone-523> ffmpeg.git 03Clément BSsch 07master:7b80b3cef0a2: lavc/utils: merge side data after video encode.
[02:18] <cone-523> ffmpeg.git 03Clément BSsch 07master:13478b270adf: gif: use only one graphic control extension block per image.
[02:18] <cone-523> ffmpeg.git 03Clément BSsch 07master:7e57adb464a9: lavf/gif: fix timing.
[02:23] <ubitux> that's so much better with proper timestamps..
[02:26] <durandal_1707> ubitux: could you make curve line stuff in curves filter available to other filters, it could be nice addition to afade
[02:27] <ubitux> you mean the dot interpolation?
[02:27] <ubitux> how would that help afade?
[02:28] <ubitux> feel free to take it out; that's hard to factorize code when you don't use it directly
[02:28] <durandal_1707> ubitux: to set custom fade curves, obviously
[02:29] <ubitux> feel free to export and adapt to your needs :p
[02:57] <cone-523> ffmpeg.git 03Paul B Mahol 07master:b63f7b3ac7b9: lavfi/afade: use AV_OPT_TYPE_DURATION
[02:59] <durandal_1707> i dont get this guy that use pullup on source that is not telecined
[03:22] <durandal_1707> colorchannelmixer=.8:.2:0:0:.8:.2:0:0:.8:.2:0:0,colorbalance=rs=.1:gs=-.01:bs=-.1:rm=0.05:gm=-.005:bm=-.05 
[10:05] <cehoyos> ubitux: Ping
[10:06] <ubitux> cehoyos: yeah i know
[10:06] <ubitux> user-agent? :)
[10:06] <cehoyos> No
[10:06] <cehoyos> I believe what we do now is better
[10:06] <cehoyos> (But I have no strong opinion)
[10:06] <cehoyos> I looked again into the samples from the user who cannot switch from MEncoder because of pullup
[10:07] <cehoyos> As said the samples only contain progressive frames, no interlaced ones
[10:07] <cehoyos> The transmission is 60000/1001 frames, the original source 24000/1001
[10:07] <cehoyos> The problem is - afaict - that I (he) cannot tell the decimate filter to drop three of five
[10:07] <cehoyos> Works somehow with mpdecimate
[10:08] <cehoyos> Is it possible/difficult to change decimate in this regard?
[10:08] <cehoyos> https://dl.dropboxusercontent.com/u/42718751/Blockyclip_mpg.zip
[10:09] <nevcairiel> thats a pretty normal 3:2 telecine which was deinterlaced to 60 fps, its a common pattern that should work somehow
[10:09] <ubitux> cehoyos: drop 2 you mean?
[10:09] <cehoyos> But you agree that fieldmatch will not help, don't you?
[10:10] <cehoyos> I think three (but I may be wrong)
[10:10] <cehoyos> After all, it is 3:2:3:2:3:2:3:2
[10:10] <cehoyos> so three of five frames are just duplicates
[10:11] <ubitux> sorry, 1 out of 5
[10:11] <ubitux> 2 fields in 5 frames
[10:11] <cehoyos> That will make 60fps -> 48 fps
[10:11] <nevcairiel> its a 6:4 pattern really
[10:12] <cehoyos> Wouldn't 6:4 mean that when looking at six frames, four contain information?
[10:13] <cehoyos> But with the given sample, two frames out of five are useful
[10:13] <nevcairiel> its about fields
[10:13] <cehoyos> But the given sample shows no fields, only frames
[10:13] <nevcairiel> because it was deinterlaced
[10:14] <cehoyos> I am saying that is not true but how does this help?
[10:14] <nevcairiel> telecining 24/1.001 produces 29.97 fps, if you then deinterlace that with a video-mode deinterlacer, you get 59.94 fps
[10:14] <cehoyos> So the solution is to interlace it first, inverse telecine then?
[10:15] <nevcairiel> no, a proper ivtc can still process that pattern
[10:15] <cehoyos> It appears pullup does
[10:15] <cehoyos> (That is what the user claims, I didn't test)
[10:16] <ubitux> cehoyos: it looks like a decimate is indeed what the user is looking for; problem with current decimate is that specifying the pattern is kind of limited
[10:17] <ubitux> you can say to drop 1 out of M
[10:17] <ubitux> and not N out of M
[10:18] <cehoyos> 1 out of 2,5 ?
[10:18] <ubitux> yes not possible
[10:18] <ubitux> (yet)
[10:20] <ubitux> cehoyos: vid.stab is more advanced than our current implementation
[10:44] <cehoyos> ubitux: As said, an external deshake library is bound to rot
[10:44] <cehoyos> And why should filters not buffer a few frames?? And how would that work in an external library if the API does not support it???
[12:08] <durandal_1707> ubitux: when to fix TODOs in silencedetect?
[12:09] <ubitux> i don't plan to
[12:11] <cone-8> ffmpeg.git 03Clément BSsch 07master:a7c5b7a6764b: lavf/gif: correct the delay after the first frame.
[12:11] <cone-8> ffmpeg.git 03Clément BSsch 07master:6189ff3679ab: lavf/gif: avoid forward declaration.
[12:12] <ubitux> ./ffmpeg -i ~/samples/big_buck_bunny_1080p_h264.mov -vf "select='gt(scene,0.4)', scale=240:-1:flags=+error_diffusion, setpts=0.8*N/TB" -frames:v 10 -y bbb-preview.gif
[12:12] <ubitux> ’ http://lucy.pkh.me/bbb-preview.gif
[12:27] <durandal_1707> what is difference?
[13:09] <ubitux> durandal_1707: delay between first and last frame
[13:09] <ubitux> actually, delay between first and second
[13:18] <cone-8> ffmpeg.git 03Clément BSsch 07master:380cfce2b213: lavc: add AV_CODEC_PROP_TEXT_SUB.
[13:28] <cone-8> ffmpeg.git 03Christophe Gisquet 07master:76c727738512: x86: sbrdsp: implement SSE2 hf_apply_noise
[14:24] <Snaggle> highgod: ping
[14:26] <ubitux> lol
[14:47] <cone-8> ffmpeg.git 03Clément BSsch 07master:30e6a02c23ee: lavf/http: remove duplicated dec flag.
[14:48] <ubitux> mime_type and cookies have no opt flags
[14:48] <ubitux> not a problem?
[14:55] <cone-8> ffmpeg.git 03Michael Niedermayer 07master:42bcc4082d64: avcodec/mpegvideo_motion: Use a field from the current frame if the last is unavailable in DMV & 16x8
[15:35] <ubitux> -rw-r--r-- 1 ubitux ubitux 378K Apr 19 15:32 bbb2.gif
[15:35] <ubitux> -rw-r--r-- 1 ubitux ubitux 1.1M Apr 19 15:32 bbb.gif
[15:35] <ubitux> much better. :)
[16:29] <ffmpeg|saste> ok we are #3 in the gsoc queue, will start in ~90 minutes
[16:30] <ubitux> do we need to /nick?
[16:30] <ubitux> i don't remember
[16:31] <ffmpeg|saste> ubitux, yes if it works like the past year
[16:31] <ffmpeg|saste> so that you can participate to the discussion, in case you have questions
[16:32] <ubitux> ok
[16:32] <ubitux> only need to /nick?
[16:32] <ffmpeg|saste> then you may need to ask kblin, since i already queued my nick
[16:32] <ffmpeg|saste> ux??
[16:34] Action: Compn just going to watch
[16:34] <ffmpeg|saste> i don't think you need to queue yourself, since the "org" was already queued
[16:35] <ffmpeg|ux> ffmpeg|saste: okay, too late
[16:35] <ffmpeg|ux> "ux" because i was lazy
[16:38] <iive> so, ffmpeg have been rejected again?
[16:38] <ffmpeg|ux> iive: yes, but libav and vlc as well
[16:39] <JEEB> well, I'm not sure if VLC even applies nowadays
[16:39] <iive> huh
[16:39] <JEEB> ever since the '11 or so code-in thing
[16:39] <JEEB> where they didn't like that asm tasks for x264 were being added
[16:39] <JEEB> s/VLC/videolan/
[16:40] <funman> don't remember if we applied this year, feepk might know
[16:41] <funman> we did and were rejected
[16:42] <funman> twitter.com is a more important open source project anyway
[16:42] <ffmpeg|saste> funman, why don't you do a vlc-soc?
[16:42] <ffmpeg|saste> we should as well, but we lack the money ;-)
[16:42] <funman> we already have bounties
[16:43] <funman> it's a nice idea though, just need someone to manage it
[16:45] <ffmpeg|saste> you guys have a serious identity problem
[16:47] <j-b> :)
[16:54] <iive> where is the list with the proposed tasks?
[16:54] <cptspiff> http://wiki.multimedia.cx/index.php?title=FFmpeg_Summer_of_Code_2013
[16:55] <cptspiff> was just reading it so..
[17:50] <ffmpeg|saste> do you have specific questions w.r.t. gsoc rejection?
[17:51] <ffmpeg|saste> i'm just going to ask for generic reasons, and for advices
[17:51] <ffmpeg|saste> i expect they will say it was a problem of space
[17:55] <ffmpeg|ux> i've no particular question
[17:59] <ffmpeg|ux> starting in 2 min
[17:59] <ffmpeg|ux> (not ffmpeg though)
[18:05] <ffmpeg|ux> ffmpeg|saste: maybe why they say that the tasks page is the most important, and so what was wrong with our, given the efforts we put into it this year
[18:05] <ffmpeg|ux> ...and given that we -afaik- honored all the requests from previously
[18:07] <ffmpeg|saste> ffmpeg|ux,  "and we particularly were aggressive about rejected orgs that had participated before"
[18:07] <ffmpeg|saste> i can't interpret this...
[18:07] <cone-723> ffmpeg.git 03Michael Niedermayer 07master:833bc6678f7f: avformat: change seeking index consistency asserts to av_asserts
[18:07] <ffmpeg|saste> so we participated before, and we also have been rejected
[18:12] <ffmpeg|ux> rockbox rejected too for no particular reason it seems
[18:12] <ffmpeg|ux> they didn't like at all a/v this year
[18:13] <ffmpeg|saste> musicbrainz?
[18:13] <ffmpeg|saste> even if the "multimedia" part is secondary in that case
[18:15] <BBB-work> is there a filter to overlay video temporally (instead of spatially, like what overlay does)
[18:16] <ffmpeg|ux> overlay like overlay?
[18:16] <ffmpeg|saste> BBB-work, interleave?
[18:16] <BBB-work> e.g. I have ffmpeg -i file1 -i file2 -filter_complex "[0:0][1:0]somefilter" outfile
[18:16] <BBB-work> and then some time settings like frame offset and number of frames for each of the two inputs
[18:18] <ffmpeg|saste> BBB-work, that is you want to display the streams one after the other?
[18:18] <ffmpeg|saste> in that case there is concat, or interleave if you want to mix them up
[18:18] <ffmpeg|saste> (interleave is still a patch)
[18:20] <BBB-work> concat only does whole streams?
[18:20] <ffmpeg|saste> yes
[18:20] <BBB-work> I need something that mixes like 100 frames of stream 1, then the second 100 frames of stream 2, etc.
[18:20] <BBB-work> I don't think concat does that?
[18:20] <ffmpeg|ux> i have a pending enable patchset
[18:20] <ffmpeg|ux> to add timeline to various filters
[18:20] <ffmpeg|ux> it could be done for concat as well
[18:20] <ffmpeg|ux> eventually
[18:21] <BBB-work> so I suppose for now I have to use head and tail and bash?
[18:21] <ffmpeg|ux> you could use select and concat
[18:21] <ffmpeg|ux> or the trim filter anton is working on instead of select
[18:21] <ffmpeg|saste> it should be possible, just it would be ugly right now
[18:22] <BBB-work> is select documented somewhere?
[18:22] <ffmpeg|saste> BBB-work, of course :)
[18:22] <ffmpeg|saste> ffmpeg-filters(1)
[18:23] <ffmpeg|ux> problem with select might be the timestamps
[18:23] <ffmpeg|ux> they won't be changed, so concat may have trouble
[18:24] <BBB-work> I can output to .yuv
[18:24] <BBB-work> I don't see a start_offset_n_frames like property
[18:24] <BBB-work> I guess 'n' is number of frames to select?
[18:24] <ffmpeg|ux> select='between(N,100,200)'
[18:24] <BBB-work> ah
[18:24] <BBB-work> ok
[18:24] <BBB-work> does that destroy the input?
[18:25] <ffmpeg|ux> destroy?
[18:25] <ffmpeg|ux> BBB-work: http://ffmpeg.org/ffmpeg-filters.html#Examples-36
[18:26] <BBB-work> e.g. if I do ffmpeg -i file -filter_complex "[0:0]select=between(n,0,100)[a];[0:0]select=between(n,200,300)[b];[a][b]concat"
[18:26] <BBB-work> does that work?
[18:26] <BBB-work> or do I need to specify -i twice then?
[18:26] <ffmpeg|ux> [0:0] split [a][b]; [a] select=...; [b] select=...
[18:26] <BBB-work> oh that works too I s'pose
[18:26] <BBB-work> ok
[18:33] <BBB-work> [e2]select=between(n,842,901)[e4] ?
[18:33] <BBB-work> [Parsed_select_5 @ 0x7fb6f1417880] [Eval @ 0x7fff572adf30] Missing ')' or too many args in 'between(n'
[18:33] <BBB-work> [Parsed_select_5 @ 0x7fb6f1417880] Error while parsing expression 'between(n'
[18:33] <BBB-work> [AVFilterGraph @ 0x7fb6f1414960] Error initializing filter 'select' with args 'between(n'
[18:33] <BBB-work> Error configuring filters.
[18:34] <ffmpeg|saste> select='between(n,842,901)'[e4]
[18:34] <ffmpeg|ux> ^ assuming double quote around the filtergraph
[18:34] <ffmpeg|saste> syntax can be a bitch, since "," is used to separate filter descriptions
[18:34] <ffmpeg|ux> with a sane shell
[18:34] Action: ffmpeg|saste is sorry for the sessism
[18:40] <ffmpeg|ux> hey
[18:40] <ffmpeg|ux> who is up to write a filtergraph web generator
[18:40] <ffmpeg|ux> using stuff like http://jsplumbtoolkit.com/jquery/demo.html ?
[18:40] <ffmpeg|ux> we can auto-generate all the forms from the filters AVOptions
[18:41] <ffmpeg|ux> and then just have a static html page updating a filtergraph string to copy paste in ffmpeg
[18:41] <ffmpeg|ux> (or the other way around)
[18:44] <BBB-work> [Parsed_select_5 @ 0x7f9d824170c0] [Eval @ 0x7fff5a853f20] Unknown function in 'between(n,0,120)'
[18:44] <ffmpeg|ux> are you up to date?
[18:44] <ffmpeg|ux> between() was added a few weeks ago
[18:45] <ffmpeg|ux> otherwise there is the long form
[18:45] <ffmpeg|ux> gt(n,0)*lt(n,120)
[18:46] <BBB-work> let me rebase
[18:53] <BBB-work> concat only takes 2 inputs?
[18:53] <BBB-work> what if I want to concat 3?
[18:53] <BBB-work> [a][b]concat,[c]concat?
[18:54] <BBB-work> (trying to prevent having to go [a][b]concat[z];[z][c]concat[y]...)
[18:54] <ffmpeg|ux> https://www.ffmpeg.org/ffmpeg-filters.html#concat
[18:54] <ffmpeg|ux> you can have N inputs
[18:54] <BBB-work> oh n=3
[18:54] <BBB-work> ok
[19:00] <BBB-work> omg these commandlines get very long very quickly
[19:00] <BBB-work> I guess you need to script this after all
[19:01] <ffmpeg|saste> mostly, yes
[19:04] <Compn> funman : so howd that go 
[19:04] <funman> pretty wtf
[19:04] <funman> apparently we were rejected for a private reason
[19:05] <funman> waiting for feepk to tell us more
[19:05] <Compn> awesome private reason :D
[19:05] <Compn> more secrets
[19:06] <ffmpeg|ux> :D
[19:06] <Compn> ffmpeg|ux : did you queue up ? what number is ffmpeg ?
[19:06] <ffmpeg|ux> http://kblin.org/queue.txt
[19:06] <ffmpeg|ux> soon
[19:06] <JEEB> well, videolan fell out of favor with that one year's google code-in :V so I wouldn't be surprised if they still had videolan blacklisted
[19:06] <funman> see #videolan
[19:06] <ffmpeg|ux> maybe that's because of the youtube lua scripts
[19:07] <Compn> funman : i wasnt in there, did i miss it ?
[19:12] <iive> well, nice questions from VideoLan team :)
[19:13] <iive> but yeh. this thing atm is just PR event.
[19:14] <Compn> wow
[19:14] <Compn> ffmpeg|ux / ffmpeg|saste : so how will this change your question to gsoc ? :P
[19:17] <ffmpeg|durandal_> there is only 4 there we need 10 more
[19:20] <ffmpeg|ux> we're next
[19:21] <Compn> 10 more what ?
[19:21] <ffmpeg|durandal_> ffmpeg| in channel
[19:21] <Compn> dont think it will help :D
[19:23] <funman> /nick ffmpeg|funman
[19:23] <ffmpeg|saste> ffmpeg is going to be discussed on gsoc
[19:23] <iive> have even google changed their mind after this kind of meeting? 
[19:23] <Compn> its not a meeting
[19:23] <funman> http://kblin.org/queue.txt after fairmat?
[19:23] <Compn> its a rejection reason session
[19:23] <Compn> iive ^
[19:24] <Compn> (i still dont get why they dont just email the reason to people)
[19:24] <funman> they invent it on the fly
[19:27] <ffmpeg|ux> to give the illusion of a discussion
[19:30] <iive> yep
[19:31] <Compn> well good luck you gsoc guys :)
[19:31] <Compn> oh one more :D
[19:39] <ffmpeg|saste> here we go
[19:39] <iive> ffmpeg|saste: you came to ffmpeg as gsoc student, didn't you?
[19:39] <ffmpeg|saste> no
[19:40] <ffmpeg|ux> hey
[19:40] <ffmpeg|ux> and me! :(
[19:40] <iive> you did? I though you were mplayer developer ;)
[19:40] <Compn> another secret reason 
[19:45] <ffmpeg|ux> iive: no sorry i was OT (i wasn't voiced on #gsoc)
[19:51] <saste> well basically she said there is nothing specifically wrong with our application
[19:52] <saste> but there are problems related to the issues with the other involved projects
[19:52] <saste> in other words she suggests that we don't hate each other :)
[19:52] <Compn> vlc torpedoed us , and we torpedoed vlc ?
[19:52] <Compn> or libav ? x264 ? 
[19:53] <ubitux> http://b.pkh.me/gsoc-2013
[19:57] <burek> ffmpeg got rejected for gsoc?
[19:58] <Compn> yes
[19:59] <burek> and libav got accepted?
[19:59] <Compn> no
[20:00] <Compn> vlc was rejected as well
[20:00] <burek> why
[20:00] <ubitux> drama
[20:01] <Compn> super drama
[20:01] <burek> any link?
[20:01] <ubitux> that's the worst kind of drama
[20:01] <ubitux> everything in secret
[20:01] <iive> saste: i hope you told them ffmpeg is happy to work with any project.
[20:04] <burek> what was the benefit of participating in that thing so far anyway?
[20:05] <iive> burek: a lot of the students remain developers after gsoc
[20:05] <Compn> burek : you mean ffmpeg participation in ffmpeg? lots of demuxers/decoders/etc
[20:06] <Compn> you can see the list on the multimedia wiki
[20:06] <burek> yes, i mean what was ffmpeg's interest to participate in that
[20:06] <Compn> getting people to work on code that was useful to ffmpeg ?
[20:06] <nevcairiel> the biggest interest is that students who participate like what they were doing, and stick around after it ended, good recruiting of new developers
[20:06] <Compn> free money for that
[20:06] <Compn> kostya stuck around after it
[20:07] <Compn> i'm sure there were others i'm forgetting...
[20:08] <burek> i'm asking (in the first place) to get the idea what was the goal for ffmpeg and to try to see if there is another way to reach that goal too
[20:09] <Compn> maybe we dont understand your question
[20:09] <Compn> could you rephrase it ?
[20:09] <burek> well you did answer it pretty good.. the question was something like why did ffmpeg participate in gsoc so far, what was the goal for such an intention
[20:09] <Compn> the goal of ffmpeg is to get more devels, and to get more decoders and demuxers and other pieces of code that need to be done on our todo list
[20:10] <Compn> my brain is melting today, nevermind
[20:10] <burek> is there anything we could do to make students interested in the ffmpeg project, beside participating in gsoc?
[20:10] <Compn> offer bounties (cash) for code
[20:11] <Compn> contact universities / programming classes to advertise such bounties
[20:11] <burek> do we have funds for such thing?
[20:11] <Compn> there are some in SPI 
[20:12] <Compn> from time to time, there are also corporate sponsors who have offered some money in the past.
[20:12] <saste> people, i'm going to leave now, i'll send a mail on ML later tonight or tomorrow
[20:12] <Compn> vlc also has some funds, and has thought about doing their own version of a soc
[20:12] <saste> stay tuned ;-)
[20:12] <Compn> saste : ok, thanks for handling it :)
[20:12] <burek> saste o/
[20:13] <burek> Compn, that's a great idea from vlc
[20:13] <burek> and exactly what i was going for :)
[20:13] <Compn> they might have done it before, i dont recall
[20:13] <iive> ffmtech was founded for that reason too.
[20:13] <iive> anybody know what is going on with it?
[20:14] <burek> would it mean something if ffmpeg could provide some kind of a certificate or recommendation for graduate students' future employers or so?
[20:14] <Compn> ask reimar, hes on it
[20:14] <Compn> burek  : i think instead, a student would put it on his/her resume
[20:14] <Compn> or 'cv'
[20:15] <Compn> that he/she worked on ffmpeg and had code submitted
[20:15] <burek> in short, a student works on ffmpeg for a while, solving some tasks and in return he gets some kind of recommendation from the core ffmpeg developers, which might help him get a good job or so
[20:15] <burek> that way even bounties might be avoided if needed
[20:16] <Compn> its doable
[20:16] <funman> btw videolan also has some funds which could go to a ffmpeg managed project
[20:17] <burek> cv also, yes (im not sure if the recommendation is the right word, but when you are seeking for a job and you have some references from important people that tell something nice about you and your work)
[20:17] <funman> also there is http://wiki.videolan.org/Bounties#Libavcodec_bugs
[20:18] <funman> and btw putting ffmpeg on my resume gave me a job directly, even bypassing the interview :)
[20:18] <burek> great :)
[20:18] <Compn> burek : recommendation/references are correct terms. yes
[20:18] <burek> so, why can't we make our own soc :)
[20:18] <funman> "major FFmpeg contributor" (i fixed a typo in a comment of an obscure feature)
[20:18] <burek> :)
[20:18] <Compn> burek : we can, just google money is also good too :)
[20:18] <burek> funman +1 :)
[20:19] <Compn> plus google has more advertising for gsoc i think
[20:19] <Compn> how do they get students to join ?
[20:19] <Compn> where do they advertise ?
[20:19] <burek> well, we cant compare with google, that's for sure, but we can try something similar
[20:21] <burek> for example, ffmpeg could talk to some multimedia-based companies that are seeking employees specializing in multimedia and those companies could be sponsors for that kind of a thing, benefiting from future skilled employees or something
[20:21] <Compn> could be 
[20:21] <Compn> might be hard to find which companies use ffmpeg and which do not
[20:23] <burek> we could also add some more documents/tutorials for the developers, which might help students get into the coding quickly and focus on the specific problems instead of wasting time on unimportant things, etc.
[20:23] <burek> less important*
[21:09] <durandal_1707> ubitux: trans saves that much with gif?
[21:10] <ubitux> with this case, it seems to be be worse with some other
[21:11] <ubitux> in this case there is not much scene change so it's very efficient
[21:11] <ubitux> i'm trying to write heuristics to select or no transparency
[21:11] <ubitux> since it may increase the size in some situation
[21:13] <durandal_1707> hmm. its for !pal8 only?
[21:14] <ubitux> yes most of the code is for !pal8
[21:14] <durandal_1707> why?
[21:14] <ubitux> that's the default case, and it complexifies the code
[21:14] <ubitux> crop is also for !pal8
[21:15] <ubitux> for now at least
[21:15] <ubitux> it can be improved later, but i'm trying to improve the simple cases
[21:20] <durandal_1707> well you could do it all and then send single big patch.....
[21:23] <durandal_1707> michaelni: what is gonna happen with s302 encoder patch? i could pick it up...
[21:23] <durandal_1707> plan to do similar with dts encoder (maybe even als)
[21:24] <michaelni> durandal_1707, feel free to pick it up
[21:41] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:505642f18276: mp3dec: fallback to generic seeking when a TOC is not present
[21:41] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:194d0399a29c: Merge commit '505642f18276aed03278ac91b1f334ea888eac6a'
[21:53] <cone-429> ffmpeg.git 03Anton Khirnov 07master:a7f46586bf47: ff_get_buffer(): allocate the frame for max(coded,display) dimensions
[21:53] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:a8b05dde0d88: Merge commit 'a7f46586bf47174b5fa00a905b767b1781ec8b72'
[21:53] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:05b2c998c7bf: avcodec: Fix lowres handling in buffer allocation
[22:09] <kierank> durandal_1707: you need a way of signalling the video frame rate
[22:10] <kierank> because the point of s302m is to have audio frames the same size as video
[22:15] <durandal_1707> you mean nubmer of encoded samples is time single frame is displayed?
[22:22] <kierank> correct
[22:30] <durandal_1707> and how would i do that?, with extradata?
[22:39] <burek> does anyone understand this question? :) http://ffmpeg.gusari.org/viewtopic.php?f=11&t=897
[22:51] <cone-429> ffmpeg.git 03Anton Khirnov 07master:5e83d9aced2f: h264: fully support cropping.
[22:51] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:f4b05cd8415a: Merge commit '5e83d9aced2fc2b2e1360452794c58aba55d497c'
[23:01] <cone-429> ffmpeg.git 03Diego Biurrun 07master:3eec1d41b4a9: fate: Reuse VREF and AREF variables where appropriate
[23:01] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:4cb59ad84b97: Merge commit '3eec1d41b4a947ba497f528e68da14a8befb85b9'
[23:17] <durandal_1707> michaelni: can i access video stream fps from encoder?
[23:18] <durandal_1707> otherwise perhaps sending side datat thing from muxer to encoder...
[23:18] <durandal_1707> using private option is imho hack
[23:18] <cone-429> ffmpeg.git 03Diego Biurrun 07master:b357f1d1d3dd: fate: Split audio filters into their own separate file
[23:18] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:6c5989cce9e3: Merge commit 'b357f1d1d3dd1d6672df6a397a31be425d31b6d1'
[23:22] <michaelni> durandal_1707, in what case (where timebase != 1/fps) does the encoder need the frame rate ?
[23:22] <durandal_1707> michaelni: so i just need timebase ?
[23:27] <ubitux> well
[23:27] <ubitux> i can't come up with good heuristics
[23:27] <ubitux> even trying both transparent and not transparent doesn't provide much interesting results
[23:28] <michaelni> durandal_1707, dunno maybe
[23:28] <michaelni> i dont know why you need the frame rate ...
[23:30] <durandal_1707> michaelni: nubmer of samples sent to audio encoder should be same as it takes it to display video frame
[23:31] <durandal_1707> and its variable for some fps, like 29,97 comes in sequence
[23:32] <durandal_1707> of 1602, 1601, 1602, 1601, 1602 samples
[23:32] <cone-429> ffmpeg.git 03Diego Biurrun 07master:a77aed142cad: fate: Rename video filters file and add separate video filters target
[23:32] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:e07e8882ae92: Merge commit 'a77aed142cade3beb0cd43f879d4f47daae23b8a'
[23:37] <mateo`> durandal_1707: reminds me of mxf :)
[23:37] <durandal_1707> it would be ideal if its handled internally
[23:38] <michaelni> if needed we could add a frame_rate field to AVCodecContext if theres nothing there yet that can be used
[23:39] <durandal_1707> michaelni: but there is already VARIABLE_FRAME_SIZE flag for audio encoders
[23:40] <durandal_1707> so it make sense to do it internally.....
[23:43] <michaelni> audio ? i thought its about video
[23:43] <michaelni> if you want to make audio frames match video frames (and its pcm) then probably easiest to do it i the muxer
[23:43] <durandal_1707> its s320m
[23:44] <durandal_1707> encoder and ts muxer
[23:44] <cone-429> ffmpeg.git 03Diego Biurrun 07master:bc7fbd36ea85: fate: cosmetics: More sensible order for entries in video filter file
[23:44] <cone-429> ffmpeg.git 03Michael Niedermayer 07master:8f116bf71b23: Merge remote-tracking branch 'qatar/master'
[00:00] --- Sat Apr 20 2013


More information about the Ffmpeg-devel-irc mailing list