Ffmpeg-devel-irc

Download

ffmpeg-devel-irc@ffmpeg.org

March 2014

1 participants
62 discussions

ffmpeg-devel.log.20140228
by burek 01 Mar '14

01 Mar '14

[02:06] <cone-510> ffmpeg.git 03James Almer 07master:2163a40a4693: x86/imdct36: use sse3 instructions in the last BUTTERF step when possible [02:13] <cone-510> ffmpeg.git 03Diego Biurrun 07master:f2408ec9d752: Give IDCT matrix transpose macro a more descriptive name [02:13] <cone-510> ffmpeg.git 03Michael Niedermayer 07master:5c634cbeb7b4: Merge remote-tracking branch 'qatar/master' [02:26] <cone-510> ffmpeg.git 03Lukasz Marek 07master:0025f130059d: lavf/libssh: fix seek to nagative position [02:26] <cone-510> ffmpeg.git 03Lukasz Marek 07master:9f4b55ef49c1: lavf/ftp: fix seek to nagative position [02:26] <cone-510> ffmpeg.git 03Michael Niedermayer 07master:e8e91a32e5ce: Merge remote-tracking branch 'lukaszmluki/master' [03:39] <cone-510> ffmpeg.git 03Marton Balint 07master:35ead1f281ad: lavfi/frei0r: search plugins in lib64 dirs on all architectures [03:39] <cone-510> ffmpeg.git 03Marton Balint 07master:ead0dd3cf6e5: lavfi/frei0r: factorize plugin loading with static path [03:39] <cone-510> ffmpeg.git 03Michael Niedermayer 07master:669043db811a: Merge remote-tracking branch 'cus/stable' [03:48] <BBB> plepere: yes you can combine them into a single 2d filter (mru did that for vp8 arm/neon), but I'm lazy and this is less code [03:48] <BBB> plepere: it's mostly just a question of "how lazy am I" [03:49] <BBB> plepere: as for fixed-width no speed-up, show me the code, would love to dig into this [03:49] <smooochysix4> hello everyone [03:50] <smooochysix4> I need a little help with ffmpeg... [03:52] <smooochysix4> I want to use ffmpeg on Linux Mint, but I want to make sure that it the codecs can be used by all multimedia programs like Totem. Do I have to compile it as static or shared? [04:02] <relaxed> smooochysix4: user questions --> #ffmpeg [04:09] <klaxa> * smooochysix4 has quit (Quit: Leaving) [04:09] <klaxa> <relaxed> smooochysix4: user questions --> #ffmpeg [04:09] <klaxa> little too late [05:07] <relaxed> klaxa: ha, my client is configured to ignore parts/joins [05:07] <klaxa> yeah i think we had something similar in #ffmpeg with a user that frequently disconnected due to bad connectivity of their client [09:06] <plepere> BBB : I've got all previous functions with fixed-width. I'm refactoring to remove the "random registers" by named ones (through cglobal) [09:07] <plepere> I think I'll submit another patch later once I've got all this done, so everyone can check it out [10:44] <ubitux> meh, is diego moving dsp code from proresdsp to the proresenc itself? [10:44] <ubitux> ah, arguments are "just cruft" and "unlikely to get optimized" [12:47] <BBB> plepere: cool [12:48] <BBB> ubitux: just don't merge the patches; I mean, diego doesn't know anything about code [12:48] <BBB> his most-used keys in his code editor are spacebar, backspace and enter [12:49] <durandal_1707> probably using emacs [12:50] <BBB> if you're not using the features of the ice, it doesn't really matter which one you use [12:50] <BBB> s/ice/ide/ [14:35] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:f3eef02746ef: avcodec/msvideo1: Fix palette in case of seek before decode [14:47] <plepere> should I squash all my commits in a single patch ? it's quite similar to the previous patches I've done. The significant changes are in the .asm file [14:50] <plepere> I think I'll copy my modified files somewhere, reset the branch and paste the files, giving clean commit messages and all. :) [15:03] <iive> plepere: you can easily merge commits in git. `git rebase -i <past commit to start with>` -i stands for interactive. [15:05] <plepere> iive : yes, but I haven't really followed the "commit everytime you change a file" practice, so I have commits over multiple files and all... [15:05] <iive> but generally, if patches are standalone, it is better to be separate. if you can't use one without the other... then consider merging them. [15:06] <plepere> iive : I'm considering doing 2 patches : 1 to modify C functions, 1 to add all the ASM. [15:06] <plepere> I'm adding HEVC ASM functions [15:07] <iive> makes sense. [15:08] <iive> there was an interactive git command that could let you split a portion of a commit into a separate one. I just don't have it in my hints. [15:09] <plepere> i've read about it, yes [15:09] <plepere> but i'ts easier for me to just copy-paste the couple of files. :) [15:09] <iive> but since it does line by line, it could take some time :) [15:09] <smarter> git reset --soft HEAD somefile.c [15:09] <smarter> git commit --amend [15:10] <iive> the soft reset would remove the commits, but leave the files in place. [15:11] <iive> the amend works only on the previous commit. [15:12] <iive> he said he have many commits on it. [15:12] <smarter> ah right, I mean git reset whatevercommitdoesnthavethefile somefile.c && git commit --amend [15:13] <smarter> ah, I thought he had one commit he wanted to split into two [15:13] <smarter> easier to squash everything in one commit first [15:13] <plepere> i think your technique is the cleaner version [15:13] <plepere> :p [15:13] <iive> plepere: btw, backup your history, just in case. [15:13] <nevcairiel> or just reset back to your branch-off and redo all the commits manually, which is pretty much the same as copy/pasting the files :p [15:15] <ubitux> plepere: make sure it builds & works after each commit [15:15] <ubitux> also, git add -p is your friend [15:16] <plepere> it works, I'm now doing the commit message for the ASM. :) [15:17] <J_Darnley> That very useful -p switch is also available with commit, checkout, and reset [15:18] <plepere> going to send a mail. :) [15:24] <plepere> it's gone ! [15:25] <plepere> :D [15:27] <ubitux> wow that's quite redundant [15:27] <ubitux> you should use %rep when possible [15:28] <ubitux> an probably factor out most of these [15:28] <ubitux> and actually macrotize those patterns [15:28] <ubitux> they're very similar [15:34] <plepere> %rep ? [15:34] Action: plepere googles that [15:36] <plepere> ooooh [15:36] <plepere> interesting [15:37] <plepere> but you said last time that macros made things unreadable. D': [15:38] <plepere> j/k I'll try to use the %rep and %assigns [15:40] <ubitux> plepere: huh? [15:40] <ubitux> when did i say that? [15:41] <plepere> I think it was on my previous version, where the cglobal would be out of the macro [15:43] <ubitux> yes [15:44] <ubitux> i suggested to put the cglobal into the macro [15:44] <ubitux> and the RET as well [15:44] <ubitux> but that's by no mean a request to remove the macro [15:44] <ubitux> quite the opposite [15:47] <plepere> I'm adding the %rep right now. :) [15:52] <plepere> thanks for the tip. :) [15:53] <plepere> BBB : In my patch, there are unrolled loops if you want to see the performance of fixed-width vs previous implementation [16:15] <saste> michaelni, pengvado i'll be offline in the next 30 minutes, will be back again at 16 UTC [16:15] <saste> what's the name of the gsoc channel for the rejection session? [16:16] <Compn> it was #gsoc [16:16] <Compn> i think [16:16] <Compn> yep [16:19] <ffmpeg_saste> michaelni, you going to queue our ideas page? [16:20] <ffmpeg_saste> pengvado, michaelni, ping^ [16:20] <ffmpeg_saste> the sooner we queue the better [16:22] <ubitux> wtf fraunhofer [16:22] <JEEB> yeah, they're having a field day :P [16:22] <ubitux> > What is intended by our first patch regarding the license is to provide a mechanism for us to be able to maintain this wrapper. [16:22] <ubitux> > However, after consulting legal expertise, we have determined that it would only be possible for us to contribute to FFmpeg under LGPL v2.1 only. [16:22] <ubitux> :') [16:23] <JEEB> they did note that they were sorry that they didn't make it a separate commit [16:23] <nevcairiel> i'm still not sure why anyone would even want the decoder wrapper, what features does it exclusively have? :p [16:23] <nevcairiel> 960 maybe [16:26] <ffmpeg_saste> michaelni, i'm going to queue our ideas' page right now, since I don't plan to keep waiting all day [16:26] <michaelni> ffmpeg_saste, please do [16:30] <Compn> nevcairiel : always good to have a reference to benchmark/test off of [16:30] <ffmpeg_saste> michael: uuuhhh we are #19 [16:31] <ffmpeg_saste> i'll get off and be back hopefully before 16 UTC [16:32] <ffmpeg_saste> cyl [16:44] <reynaldo> ffmpeg is by far the most known project on the gsoc rejections feedback meeting's queue thus far [16:44] <reynaldo> something to worry about I guess [16:44] <ubitux> it's been a few years we're rejected [16:45] <ubitux> i guess they wait for the libav/ffmpeg case to be solved [16:45] <ffmpeg_michaelni> btw, gsoc topic says "Change nick to orgname_nick" [16:45] <ubitux> i'm worry i won't come [16:45] <ubitux> sorry* [16:46] <ubitux> i doubt there will be anything new from previous years, and have nothing to share [16:46] <Compn> ubitux : didnt google hire people to work on ffmpeg and submit patches? so why pay us to do the same thing with students? [16:46] <Compn> google security researchers that is [16:47] <ubitux> that's unrelated [16:47] <Compn> ah [16:47] <reynaldo> Compn: the chances of retaining those are quite dim, even if you try with a good retention strategy [16:47] <reynaldo> students are supposed to be easier to elevate to regular contributor status [16:47] <reynaldo> so it still makes sense [16:48] <Compn> what are our numbers on retaining students ? :P [16:48] <reynaldo> Compn: thats another question [16:48] <Compn> fine! [16:48] <reynaldo> and another debate [16:49] <reynaldo> I saw it comming though, thats why I added "even if you try with a good retention strategy" [16:49] <reynaldo> pardon my english if that wasnt clear enough [16:49] <reynaldo> :) [16:49] <Compn> its ok, i'm kinda dense this morning :) [16:50] <reynaldo> np :) [16:50] <reynaldo> TGIF :) [16:52] <ffmpeg_michaelni> reynaldo, you probably should also change nick if you want to participate in the gsoc meeting thing, [16:53] <ffmpeg_michaelni> iam not sure but people without it might not have voice then or something [16:53] <reynaldo> on it [16:53] <ffmpeg_michaelni> dont remember how its handled [16:55] <reynaldo> ffmpeg_michaelni: there now [16:58] <Compn> all i remember is last year they spent 4 hours telling the majority of people that they just picked other projects [16:59] <Compn> would be easier to just email that rejection reason... [17:00] <reynaldo> meeting has started [17:01] <reynaldo> does someoen know why was it that xorg got rejected last year ? [17:05] <ffmpeg_saste> hi do you guys have specific answers for the gsoc admins? [17:05] <reynaldo> ffmpeg_saste: questions you mean ? [17:06] <ffmpeg_saste> reynaldo, of course [17:07] <reynaldo> ffmpeg_saste: not me, I have some general comments on favoring new/small communities not necesarily been a great idea [17:07] <reynaldo> feels a bit unfair too [17:07] <ffmpeg_saste> i think ffmpeg qualifies as "small" [17:07] <iive> maybe ask if there is any point in trying again next year... [17:08] <reynaldo> ffmpeg_saste: theres a question then, does ffmpeg qualifies as an small org ? [17:08] <ffmpeg_saste> iive, yes [17:09] <ffmpeg_saste> although i expect the same reply as the past two years, "you're welcome!" [17:09] <iive> yeh, PR [17:09] <ubitux> "it's a been a few years we're getting rejected; are we wasting our time trying?" [17:09] <reynaldo> maybe we want to get a clear answer on whether the ffmpeg/libav issue is affecting their desicions in any way [17:10] <ffmpeg_saste> i don't know if this is really related to politics or not, and we can't tell from the answers we got in the last year [17:11] <reynaldo> that supports the idea of that question been needed then [17:11] <ffmpeg_saste> pengvado, you here? [17:11] <reynaldo> did libav applied btw ? [17:11] <ffmpeg_saste> reynaldo, no, nor libav [17:11] <reynaldo> apply/ [17:12] <j-b> As I told you already [17:12] <j-b> DO NOT APPLY [17:12] <j-b> You will get rejected [17:12] <ffmpeg_saste> j-b, if it was for me, i was not going to apply, but other people in the project insisted [17:13] <j-b> /nick Cassandra [17:14] <reynaldo> I see no reason why not to tbh, even with a sure rejection in the horizon (which 'sure' being largely arguably) [17:14] <reynaldo> which/with [17:17] <ffmpeg_saste> reynaldo, ah i was convinced pengvado was the other admin... [17:17] <ffmpeg_saste> but... it's a bit annoying that they invite us to participate and then they reject us for unknown reasons [17:17] <iive> do they invite us? [17:17] <ffmpeg_saste> i would appreciate more if they said bluntly that they were not going to accept us, and save us some time [17:18] <reynaldo> loren? I didnt see him involved on any of the discussions for the applying process [17:18] <ffmpeg_saste> iive, no, why should they? [17:18] <ubitux> well didn't they actually said something like "please try again next year!" [17:18] <iive> because you said " it's a bit annoying that they invite us to participate"... and ... [17:20] <j-b> No, it's even worse than that [17:20] <j-b> They told you waht to do to get accepted the next year [17:20] <j-b> and told you how to run the comunity [17:20] <ffmpeg_saste> ubitux, past year they said, no, we should stay off for "some time" [17:20] <cone-837> ffmpeg.git 03Diego Biurrun 07master:b23650491fbd: prores: Use consistent names for DSP arch initialization functions [17:20] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:5ba164831816: Merge commit 'b23650491fbd579a4365f42bd42575afb7b53f7e' [17:21] <ubitux> ffmpeg_saste: mmh ok [17:21] <ffmpeg_saste> then i asked again this year, and i got a reply on the line "sure you're welcome, but we can't guarantee that you're going to be accepted" [17:22] <ffmpeg_saste> again, note that i decided to work as backup admin *only* because other people insisted for it [17:24] <iive> when there is uncertainty, ask for a percentage :) . [17:26] <iive> or you could say "what is the chance if we try with the same tasks as last year". [17:28] <reynaldo> hehe, I;d say that if you go with the same ideas next year you're likely failing as a project anyway [17:28] <reynaldo> that said, I doubt they'd check [17:28] <reynaldo> but upfront, I think their answer would be around those lines [17:31] <reynaldo> <carols> it sounds like an interesting project, but we would have also liked to see someone vouch for you [17:31] <reynaldo> that thing is been repeated a lot [17:31] <cone-837> ffmpeg.git 03Diego Biurrun 07master:d6acefe05862: proresenc: Drop unnecessary DCT permutation bits [17:31] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:18d870da83b1: Merge commit 'd6acefe05862af244fd5a30ae946ed507c063994' [17:31] <ffmpeg_saste> reynaldo, doesn't that apply only to *new* projects? [17:32] <reynaldo> ffmpeg_saste: havent checked, but from 6 already 60% have gotten such 'suggestion' [17:33] <iive> "vouch" in what sense? [17:33] <ffmpeg_saste> reynaldo, i'm pretty sure we as "veteran" org didn't need that, or we could have asked some people from G to vouch [17:33] <reynaldo> ffmpeg_saste: we had ppl from G vouching, at least on paper [17:34] <reynaldo> we added those to the comments at the end of the application form [17:35] <reynaldo> anyhow, from what Im seen on that meeting we were either rejected for been too senior or for the libav/ffmpeg issue, Im inclined to believe it was the former [17:35] <reynaldo> none of the rest of the negative feedback seems to apply to our application nor our ideas page [17:35] <reynaldo> should make for some fun feedback [17:36] <ffmpeg_saste> reynaldo, you are too old! -> sounds discriminatory ;) [17:37] <reynaldo> ffmpeg_saste: tell me about it [17:37] <reynaldo> :) [17:37] <reynaldo> we should sue them [17:38] <iive> yep, we may with in 10 years. :) [17:40] <JEEB> hmm, what's the opinion on http://ffmpeg.org/pipermail/ffmpeg-devel/2014-February/154078.html btw? [17:40] <JEEB> seems like dae pinged it on the 9th [17:40] <JEEB> but it hasn't been merged as far as I can see? [17:40] <JEEB> nor commented upon [17:41] <reynaldo> are we allowing fprintf(stderr now? [17:42] <JEEB> that's from the example, which seems to use them to begin with [17:42] <JEEB> internal stuff should be av_log() afaik [17:53] <Compn> j-b : i forgot why google rejected vlc the other year, was it x264 thing ? [17:54] <ubitux> because they host ffmpeg, they are likely part of the libav/ffmpeg mess [17:54] <Compn> ah [17:58] Action: Compn points fingers at j-b [17:58] <Compn> hee [17:59] <Compn> so bayer stuff is in now ? [18:00] <Compn> time to find the bayer samples and see what has to be done to hook those up [18:04] <cone-837> ffmpeg.git 03Diego Biurrun 07master:92e598a57a7c: prores: Drop DSP infrastructure for prores encoder bits [18:04] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:2e88f82a8a50: Merge commit '92e598a57a7ce4b8ac9ea56274af39f5fd888311' [18:07] <reynaldo> <carols> nope, size of the org doesn't matter [18:07] <reynaldo> thats a bit counter after she said they favored small orgs [18:07] <ffmpeg_saste> reynaldo, size doesn't matter, it's how do you use your org which matter [18:07] <ffmpeg_saste> *matters [18:07] <reynaldo> ffmpeg_saste: she said at the beginging they tried to favour small or new orgs [18:07] <reynaldo> not quoting verbatim but that was the idea [18:08] <ffmpeg_saste> reynaldo, you missed my obscene double meaning [18:08] <reynaldo> oh [18:08] <reynaldo> then yeah [18:08] <reynaldo> although I'd never know, Im only part of big orgs [18:08] <reynaldo> :) [18:08] <ffmpeg_saste> ahah [18:09] <Compn> lol [18:09] <Compn> saste: are you still in the queue [18:09] <Compn> ? [18:10] <Compn> i'm guessing yes [18:10] <reynaldo> we are, 5 to go iirc [18:10] <ffmpeg_saste> Compn -> http://kblin.org/queue.txt [18:10] <cone-837> ffmpeg.git 03Diego Biurrun 07master:a55546f48d55: proresenc: Reuse proper dsputil infrastructure for FDCT [18:10] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:810eb285e38d: Merge commit 'a55546f48d55e3d1155840541b2be5f4f8cf18ab' [18:16] <ffmpeg_saste> good, it's going fast [18:16] <Compn> maybe we're just so organized we dont need any help :P [18:18] <cone-837> ffmpeg.git 03Christophe Gisquet 07master:87ec849fe9ac: dcadec: remove scaling in lfe_interpolation_fir [18:18] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:90f674d55bd7: Merge commit '87ec849fe9acba075c843e67bcd01f256f481a18' [18:23] <Compn> Option -pix_fmt is deprecated, use -pixel_format. [18:23] <Compn> Option pixel_format not found. [18:23] <reynaldo> ffmpeg_michaelni, ffmpeg_saste: 1 to go [18:23] <Compn> ^^^ anyone want to tell me why? [18:24] <reynaldo> what ? [18:24] <Compn> about pix_fmt [18:24] <reynaldo> ah, sry [18:24] <reynaldo> not a clue [18:26] <ffmpeg_saste> reynaldo, are you going to have voice in the channel? [18:28] <reynaldo> I'd rather be you or michael tbh if there was just 1 but I think they would voice all of us [18:28] <ffmpeg_saste> reynaldo, ok [18:29] <ffmpeg_saste> i'll probably ask something in private [18:29] <reynaldo> why in private ? [18:29] <reynaldo> does it relate to size ? [18:29] <reynaldo> :) [18:30] <ffmpeg_michaelni> i hate private stuff for floss projects [18:30] <Compn> double secret probation [18:30] <ffmpeg_saste> ssh don't ask [18:30] <Compn> :P [18:30] <Compn> me too [19:14] <Compn> cant get lbyr sample to decode with bayer , might not be bayer? http://samples.ffmpeg.org/V-codecs/LBYR/Video000.avi [19:14] <Compn> could be anything [19:36] <cone-837> ffmpeg.git 03Christophe Gisquet 07master:ad507d790745: x86: dcadsp: implement SSE lfe_dir [19:36] <cone-837> ffmpeg.git 03Christophe Gisquet 07master:169243112c1e: x86: dcadsp: implement SSE lfe_dir [19:36] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:e346a5938317: Merge commit 'ad507d7907457e678900bac132122ba7be4644cb' [20:04] <cone-837> ffmpeg.git 03Christophe Gisquet 07master:57b1eb9f75b0: dcadsp: scan coefficients linearly in dca_lfe_fir [20:04] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:5333e0dd66c5: Merge commit '57b1eb9f75b04571063ddec316e290c216c114ac' [20:49] <JEEB> michaelni, any comments on http://ffmpeg.org/pipermail/ffmpeg-devel/2014-February/154078.html btw? [20:49] <JEEB> dae seemed to ping it some time ago, but it was never given any replies as far as I can see [20:54] <cone-837> ffmpeg.git 03Christophe Gisquet 07master:08e3ea60ff40: x86: synth filter float: implement SSE2 version [20:54] <cone-837> ffmpeg.git 03Christophe Gisquet 07master:2cdbcc004837: x86: synth filter float: implement SSE2 version [20:54] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:baf3adc621fe: Merge commit '08e3ea60ff4059341b74be04a428a38f7c3630b0' [21:30] <ubitux> michaelni: diego is starting to push his dsp code shuffling, i think we might want to avoid that in 2.2 [21:30] <ubitux> maybe we should branch out or something? [21:38] <cone-837> ffmpeg.git 03Janne Grunau 07master:7686afd049be: dca: factorize scaling in inverse ADPCM [21:38] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:747b0337e70b: Merge commit '7686afd049be98d18663682b92d983340fa2c305' [21:38] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:ffb7d7195be6: avcodec/dcadec: use brackets to ensure that no slow division is used [21:42] <michaelni> JEEB, replied, but i guess you already saw [21:45] <JEEB> k [21:46] <JEEB> too bad it seems like Dae is no longer interested in replying, so I will have to find the whole interaction related to this change from various logs [21:52] <cone-837> ffmpeg.git 03Christophe Gisquet 07master:4cb6964244fd: dcadec: simplify decoding of VQ high frequencies [21:52] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:fb3c33f3cd6b: Merge commit '4cb6964244fd6c099383d8b7e99731e72cc844b9' [22:13] <cone-837> ffmpeg.git 03Janne Grunau 07master:6e4009d4cdf5: arm: dcadsp: implement decode_hf as external NEON asm [22:13] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:fe6603745e93: Merge commit '6e4009d4cdf5927bdaedf58fcfc5e813b14c366b' [22:39] <ubitux> wtf, 2 news guys sending K&R reindent commits to libav-devel oO [22:40] <ubitux> http://cdn.memegenerator.net/instances/500x/46616414.jpg is this happening right now? [22:40] <wm4> yeah that's strange [22:40] <JEEBcz> even libav-devel seems to be o_O'ing at third parties doing that [22:40] <ubitux> space magnet [22:42] <wm4> maybe they saw all the diegoing and thought this was a good way to start contributing [22:53] <Compn> diegoing ? [22:53] <Compn> oh diego'ing [22:53] <Compn> dont mind me [22:53] Action: Compn goes back to bidness [22:53] <cone-837> ffmpeg.git 03Hendrik Leppkes 07master:3aca10bf762a: latm: Always reconfigure if no extradata was set previously [22:53] <cone-837> ffmpeg.git 03Michael Niedermayer 07master:d46c3b8171b3: Merge remote-tracking branch 'qatar/master' [23:58] <manu1083> Hello, is that it is possible to convert smil to avi with avconvert ? [00:00] --- Sat Mar 1 2014

1 0

ffmpeg.log.20140228
by burek 01 Mar '14

01 Mar '14

[00:04] <active8> anyone ever play woth security cam .264 backup files? [00:10] <llogan> active8: is it from one of those crappy chinese "DVR" things? [00:14] <active8> llogan. I couldn't find the name. I think it's on the bottom and the owner was cagey enough about me being in the office. The monitor was on top of it, too [00:15] <active8> the files are from 6 channels, 265 MB each. He's kept me waiting since yesterday for this cause he's lame. I had to bring a flash drive - he has no way to print out frames or give the cops a video. 7-1 will put your mug up for everyone to see as soon as the manager comes in in the AM [00:16] <active8> I get about an hour out of the conversion using the command in the paste and there's not even a mouse stirring at midnight when they're closed. [00:17] <active8> It is a DVR looking thing, though. Code key lock on the hard drive [00:21] <active8> llogan: Did you see the paste? Supposedly, errors are typical for those systems [00:27] <llogan> active8: looks like a bunch of decoding errors/messages. does your output look ok? [00:29] <active8> not bad for dead of the night - looks the same quality as what i saw on the monitor both live and playback. I suppose 265 MB might only be an hour, but at h264 (mp4?) my experience tells me that 265 MB could be a day and a half but I'd have to double check some of the tests I've done with other stuff here [00:30] <active8> like apples to oranges - slideshows from jpegs, screen caps... all at diff bitrates and fps [00:31] <llogan> i don't know what you're trying to do or what the issue is [00:32] <active8> i'm trying to convert it to an mp4 that vlc will play, with timestamps so I can do a -ss -t to get the few seconds of the crime. right now, -ss -t fails and gives me a file that doesn't even start playing in vlc [00:33] <klaxa> well for starters you could just copy the stream into an mp4 file [00:33] <klaxa> run: ffmpeg -f h264 -r 30 -i ch00000000000003-140226-000000-010000-12p00001000000400.264 -c copy ch3.mp4 [00:33] <klaxa> then you have it inside a container which should make it easier to handle [00:34] <klaxa> then you can try to play it with vlc, if that works, you can use -t and -ss on the .mp4 [00:34] <active8> Then the video goes to baltimore county breaking news fb page, frames go to posters, and this little peninsula's citizenry does the rest [00:37] <active8> but it ends at 1 AM, I need 4:54:45 PM which is there on the machine. I just don't know if it's here cause of the errors and going back to get lame ass store owner to play some more is like pulling teeth. A local restauranteuer leaves me in his office (where he stores liquor)and other stuff to fix his system while his serving wenches serve me food and beer for free, but this guy must be afraid I'll take a pencil. [00:38] <active8> Those last few "paragraphs" of errors just run and run and fill the terminal buffer untill it just ends, leaving me with an hour of recorded deadness. [00:39] <klaxa> did you try running the command i pasted? [00:39] <active8> oh sorry. I missed that. [00:42] <active8> doesn't like -c, it's -vcodec or -c:v iirc. I think I ran that earlier and got nada [00:42] <klaxa> what do you mean with "doesn't like -c" [00:42] <klaxa> -c is the generic codec parameter [00:43] <klaxa> if you pass copy to -c (as in "-c copy") it will copy all codecs [00:44] <klaxa> when in doubt, get a newer ffmpeg [00:51] <active8> llogan: it results in a vid that ends where the camera system timestamp is 12:45:12 and vlc says that is at 00:54:12 out of 01:11:55. So if it's the 30fps that we assume, the errors that were dropped "shortened the footage. This command really did the same as what I did with way less error messages, but the out file size is the same - 28.1 MB So I'd think the whole 250 - 275 MB (6 diff channel files) file would be closer to 10 hrs of recording [00:52] <klaxa> what is the filesize of the input file? [00:52] <active8> klaxa: -c yields Unrecognized option 'v:c' [00:53] <sacarasc> Did you do -v:c rather than -c:v ? [00:53] <active8> klaxa: I built this ffmpeg - oh sit, I forgot to add the path again [00:53] <active8> s/sit/shit/ oh friggin' shit! [00:54] <klaxa> did you run the exact copy i pasted? also from the message of the paste your ffmpeg version is from 2013 [00:54] <klaxa> >ffmpeg version git-2013-08-04-61af627 [00:54] <active8> yeah sacarasc. And forgot the damned path there. too. [00:56] <active8> man says use -codec:v or -c:v. Maybe I'll remember that for more than a few minutes [00:56] <active8> yeah. I ran the exact command. let me try it with my new build [00:58] <klaxa> if you ran the exact command please pastebin your command with output [00:58] <active8> vlc won't even play it. The old crappy ffmpeg that installs with vlc is the one that gave me the 28.1 MB file that worked [01:00] <klaxa> look, i want to help you, if you can follow my instructions that would bring us a lot further [01:08] <active8> running it again klaxa. I did run your command and got a blank. my command from the paste that I did ^^ up there ^^ was the same except for the copy codec ( i didn't use it) gave me a better result (camera system timestamp vs vlc time wise) and it played. -c:v and -vcodec copy work too, with less errors [01:08] <active8> here's the paste http://pastebin.com/VjCVuGVk [01:09] <klaxa> you said the file is 265 mb? [01:09] <klaxa> >frame=107905 fps=82958 q=-1.0 Lsize= 259398kB time=00:59:56.83 bitrate= 590.8kbits/s [01:09] <klaxa> as you can see, the result file is about the same size [01:09] <klaxa> your file is only 1 hour long [01:09] <active8> yes and here is the paste from above: http://pastebin.com/1LjjWsCk [01:09] <klaxa> that's why you can't go any further than 1 hour [01:09] <klaxa> your file is still only 1 hour [01:10] <active8> that's what i was starting to realize. Friggin store owner is a dickhead [01:11] <klaxa> please refrain from swearing [01:11] <active8> ok [01:11] <klaxa> this channel is publicly logged it may reflect badly on you [01:11] <active8> store owner gentleman didn't do the backup correctly [01:12] <klaxa> seems to be the case [01:13] <active8> where's the public log, anyway. is it easy to search for keywords like h264 conversion or "determine fps?" [01:14] <sacarasc> !log [01:14] Action: sacarasc shrugs. [01:14] <active8> !log [01:14] <klaxa> http://lists.ffmpeg.org/pipermail/ffmpeg-devel-irc/ [01:14] <active8> me too [01:14] <sacarasc> There! [01:15] <active8> bot needs caffeine. me too brb [01:15] <active8> BTW. FFMPEG rules - once you get the command to work for what you need [02:01] <Dark-knight> oh god... I just realized what time it is... [02:01] <Dark-knight> yesterday I calculated that producerseal gets home at 8:00 pm and gets off irc at 1:00 am [02:01] <Dark-knight> he gets home in 5 minutes [02:40] <klaxa> how do i generate documentation from the sourcecode? i want to learn how to use libavformat, but reading the documentation in the source itself isn't very effective, also it seems to be formated to be parsed [02:40] <klaxa> s/effective/efficient/ [02:47] <llogan> klaxa: make && make doc [02:47] <klaxa> alright thanks [02:47] <llogan> texi2html can be a build dependency if you want HTML formatted docs too [02:49] <klaxa> the documentation i found in doc/ so far is very sparse on libavformat though [02:49] <klaxa> as in... what libavformat is and that's it [02:51] <llogan> there's libavformat.texi but i don't know anything about it [02:51] <llogan> doc/examples might be more useful or informative [02:51] <klaxa> yeah i'm looking at that too, the functions used are not explained at all though, so i thought looking at avformat.h might clear some things up [02:52] <llogan> you can always ask in libav-user mailing list if you don't get an answer here. [02:52] Action: llogan is library ignorant [02:53] <klaxa> after writing a server that parses mp3 bytestream i thought i should stop re-inventing the wheel and not do the same thing for matroska [02:53] <klaxa> although the server is 100% posix compliant and does not rely on any external libraries [02:54] <klaxa> and matroska is a bit more sophisticated than mp3 [02:54] <klaxa> so there's that too [03:01] <klaxa> for the record, i could generate the documents i wanted in the first place by running doxygen [03:01] <klaxa> could have thought of that earlier [03:24] <funyun> hi. can anyone tell me what deinterlace i need from this sample? http://www.sendspace.com/file/n5exg3 [03:25] <julienb> Hello [03:26] <julienb> i have a problem of synchronisation video/audio with ffmpeg when i convert a mp4 video [03:26] <julienb> i tryed a lot of options but the problem persist [03:27] <julienb> at the end of the convert i have global headers: 0KB muxing overhead 0.2127... [03:28] <relaxed> funyun: -filter:v yadif should be fine. [03:29] <julienb> someone for help please [03:33] <julienb> http://pastebin.com/YyFMuSLC [03:35] <relaxed> julienb: First of all you're ffmpeg is realy old. Try this instead http://johnvansickle.com/ffmpeg/ [03:36] <julienb> relaxed:i use the last stable version for gentoo [03:38] <relaxed> stable is 1.0.8 https://packages.gentoo.org/package/media-video/ffmpeg [03:39] <relaxed> but even that's too old. Install ffmpeg-9999 [03:41] <julienb> relaxed: apart the version, do you see another thing which could make the problem in the pastbin? [03:41] <relaxed> julienb: maybe -async 1 after the input will help [03:41] <julienb> yes? [03:42] <julienb> i can test [03:49] <julienb> oh its much better [03:50] <julienb> still have global header ....muxing overhead [03:50] <julienb> but audio/video are synchronized [03:52] <relaxed> julienb: I don't think the global header message is an error. [03:52] <julienb> relaxed: why its appeared? [03:53] <relaxed> any error will appear on the last line of ffmpeg's output. [03:53] <julienb> ah ok so it's normal [03:55] <julienb> but that's strange i have to add -async 1, i don't remember read that information on ffmpeg documentation [03:56] <julienb> so its better i use async 1 for all my ffmpeg conversion right? [03:56] <relaxed> if you read what that option does it makes sense [03:59] <relaxed> and it's hard to say if you need it for all of what you're doing [03:59] <julienb> for all split i think i need it [03:59] <julienb> to avoid any problem of synchronisation [04:01] <relaxed> ok [04:02] <relaxed> but really, upgrade your ffmpeg [04:02] <julienb> yes i will do it [04:02] <julienb> thanks relaxed. [04:02] <funyun> relaxed: thank you :) [04:45] <Dark-knight> what is this? a channel for ants? [04:46] <Dark-knight> what up guys [04:51] <Nosomy> lol [11:20] <Waster> hey, how to change stream name in sdp wile streaming via RTP? [11:36] <IamTrying> I have a serious problem in one 40" screen [11:36] <IamTrying> when i play one file.mp4 i get in the screen some random "WHITE DOT" [11:37] <IamTrying> i have no idea what it is. i asked the designer to give me some more files he sent me MP4, AVI prepared from Windows PC [11:37] <IamTrying> i played them and i have same WHITE DOT, but he do not have it [11:38] <IamTrying> Can anyone please tell me how can i fix it? how can i convert the MP4 original to something which is compatible for Linux version?? for 40" screens [11:38] <IamTrying> how do i check what is the original file codec?, bitrate and frames value?? [11:43] <IamTrying> Hello????? [11:43] <IamTrying> how can i check the file existing bitrate please? [11:44] <IamTrying> ffprobe -show_streams -i "file.mp4" ?? [14:20] <vega13> hi everyone [14:20] <vega13> what's up, hope everything is allright [14:21] <vega13> can anyone tell me what's difference between the 2.1 and 1.2 release branches? [14:23] <sacarasc> vega13: On the http://ffmpeg.org/ look for the 2.0 and the 2.1 release changes. That's the difference. :D [14:23] <JEEB> vega13, when they were branched from master [14:24] <vega13> @sacarasc 2.0 and 2.1 are the same branch [14:24] <sacarasc> But 1.2 doesn't have the changes from either (except maybe bug fixes.) [14:25] <JEEB> you can see when the release branches were branched off by looking at the first tag done from it http://git.videolan.org/?p=ffmpeg.git;a=tags [14:25] <JEEB> 1.2 was branched off 11 months ago [14:25] <JEEB> anything newer than that, other than possible bug fixes, is not in there [14:25] <JEEB> 2.1 was four months ago [14:26] <JEEB> anything newer than that, other than possible bug fixes, is not there [14:29] <albator> i got a 3 months old ffmpeg, is it worth updating? [14:30] <albator> I guess jeeb answered :) [14:31] <JEEB> I recommend you trying what you generally do with it once, and unless there are some specific bugs, then move to a newer one [14:31] <JEEB> if there are bugs, report them [15:42] <gers4302> I'm making an AVFilterGraph in code. I need to apply a watermark to the video. My assumption is that I need to use the "movie" AVFilter to then overlay it. My watermark is just stored in memory, not in a file, so I'm unsure how to point the "movie" filter to use that watermark? Can someone clarify how I would do that? [18:30] <hans_henrik> how can i tell ffmpeg to only copy audio stream #2 ? [20:22] <matham_> is there a way using the ffmpeg eval api to get the duration of a video? I need to use it with setpts. My videos have incorrect framerate (29.97) and they should all be e.g. 50sec. So given duration, the real_fps is 29.97 * duration / 50. Then, "setpts=N/real_fps" would work. [20:25] <matham_> by they should all be e.g. 50s, I mean I know that the video was generated at some unknown fps for 50sec. [20:29] <ubitux> no lavfi has no knowledge of the stream duration [20:29] <ubitux> it can be anything [20:30] <ubitux> you can probably do ffprobe -v error -of flat -show_entries format=duration:stream=duration foobar.mp4 [20:30] <ubitux> in a first pass [20:30] <ubitux> to get it [20:30] <ubitux> note that in a lot of cases it can be innacurate [20:31] <ubitux> matham_: note that you have access to "FRAME_RATE" though [20:31] <ubitux> which is /sometimes/ defined :p [20:32] <matham_> ok thanks [20:32] <matham_> FRAME_RATE won't help me though, because that is the incorect one set when capturing [20:34] <matham_> but the ffprobe preprocessing seems like it should work. thanks [21:01] <reliability> hi [21:20] <PCMAD> hi i want to make a commercial program using ffmpeg [21:20] <PCMAD> what conditions apply [21:21] <average> read the license [21:28] <llogan> PCMAD: also see http://ffmpeg.org/legal.html [21:35] <PCMAD> ok thanks [21:39] <reliability> I have a seemingly strange question: Is there any part of ffmpeg which has some potential for "local" optimizations (instruction level parallelism, removing potential cache misses, etc., no multithreading though)? [21:39] <reliability> And it must be flaoting point. [22:48] <mgeary> hi folks [22:51] <mgeary> so, i'm confused about mp3 in ffmpeg. I understand that by default there's no built-in mp3 support, but i'm confused as to why when i do "-formats" i see both Decode and Encode for mp3 [23:06] <mgeary> okay, so at the moment, recompiling with LAME isn't an option, but the lame binary is on the machine. Can anyone tell me why i'm getting this: http://pastebin.com/xg6exZiU ? [23:06] <mgeary> (Warning: corrupt or unsupported AIFF format) [23:07] <mgeary> but if i do it in 2 steps (ffmpeg to aif file, and aif file thru lame) it works just fine [23:29] <kingsob> hey, I am wondering what is more imoortant for ffmpeg? memory, cpu, fast hard drive, etc? I am trying to select the best amazon ec2 instance for ffmpeg [23:30] <kingsob> right now its running on a m1.large, but I notice my macbook can do the same encoding in 1/2 the time [23:32] <Plorkyeran> video encoding is cpu-bound unless you're working with extremely large lossless or raw video [23:32] <Plorkyeran> ram beyond the minimum needed to not swap doesn't do anything useful [23:33] <average> there's some GPU-acceleration thread on the mailing list [23:33] <average> related to conversion using ffmpeg with OpenCL [23:34] <average> so it may depend on GPU as well, depending how your setup is [23:34] <Plorkyeran> x264 can sorta use opencl now, but it isn't particularly useful [23:34] <average> Plorkyeran: I would find it useful. maybe I'm wrong, why do you think it's not usefule ? [23:35] <average> *useful [23:35] <Plorkyeran> the motion estimation is the only part of the encoding process that the gpgpu computational model works for [23:36] <Plorkyeran> and the overhead of syncronizing stuff with opencl and sending data to/from the gpu just isn't that much faster than doing it on the cpu [23:37] <average> I see [23:37] <Plorkyeran> most of the work of encoding h264 is inherently serial [23:37] <average> couldn't more of it be moved to GPU ? [23:37] <Plorkyeran> not usefully [23:56] <manu1083> Hello, is that it is possible to convert smil to avi with avconvert [23:56] <manu1083> ? [00:00] --- Sat Mar 1 2014

1 0