[00:00] <saste> ubitux: "Each component is coded on a "1/(N*N)" scale." [00:00] <saste> can you explain this? [00:01] <ubitux> red: [0..255] ’ [0..N*N] [00:03] <saste> ubitux, uhm ok, with the example is much more clear [00:32] <ubitux> mmh the week just started and we already have the filter(s) of the week [00:32] <ubitux> we'll have another one this week i guess [00:32] <ubitux> from me at least, possibly mcdeint too [00:32] <ubitux> from saste :) [00:33] <saste> mcdeint is a bitch [00:35] <saste> ff_init_avcodec(void) { //we maybe should init but its kinda 1. unneeded 2. a bit inpolite from here } [00:35] <ubitux> haha [00:43] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:f82e7330af53: jpeg2000: merge mant type from j2k [00:44] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:840ee5e1a1a5: ffplay: dont forget passing lowres to private options [00:44] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:8cf57efdd784: j2k/jpeg2000: merge lowres code [00:53] <ubitux> michaelni: so you're reducing the diff by adding the missing feature from j2k to jpeg2000, from jpeg2000 to j2k, or both? [01:02] <michaelni> ubitux, yes reduce diff and when theres nothing left merge done [02:01] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:d66de3801ef5: j2kdec: merge get_tlm() from jpeg2000 [02:01] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:05fa3f68028d: j2kdec: random non functional changed from jpeg2000 [02:45] <ubitux> does anyone know a non-oriented-movie-release image comparison service like screenshotcomparison.com? [02:48] <ubitux> michaelni: i'm trying to add dithering into the vignette filter as you suggested [02:48] <ubitux> i took the dither prng from mandelbrot and used it directly [02:48] <ubitux> this is the result: http://www.screenshotcomparison.com/comparison/25755 [02:48] <ubitux> do you have any suggestion on how to improve it? [02:49] <ubitux> i'm using the same dv coeff (computed exactly like in mandelbrot) for every component, and prng updated for every y/x [02:55] <cone-561> ffmpeg.git 03Michael Niedermayer 07master:9c66da517f3e: j2kenc: Zero DWT struct on allocation [02:56] <michaelni> ubitux, ask me tomorrow i need to sleep [02:56] <ubitux> sure ok :) [02:56] <ubitux> i'll go soon as well [02:56] <ubitux> thanks :p [02:57] <burek> is there any aac test samples in ffmpeg that i can use to test something in ffmpeg? more precisely, is there any aac file with these specs: Audio: aac, 16000 Hz, mono, fltp [02:58] <ubitux> why not generate one? [02:59] <burek> hm, you are right, as usual :) [03:01] <burek> ubitux, just one more question, how to tell ffmpeg to use aevalsrc as an input [03:01] <burek> ffmpeg -i what -af aevalsrc=... [03:02] <burek> or something [03:03] <ubitux> burek: ffmpeg -f lavfi -i "aevalsrc='sin(2*PI*t*440)*exp(-4*mod(t,1))'" ... [03:03] <burek> oh great, we should put that in wiki if its not there already [03:03] <burek> thanks a lot man :) [03:03] <ubitux> there is something similar in the fancy filtering page [03:04] <ubitux> https://ffmpeg.org/trac/ffmpeg/wiki/FancyFilteringExamples#histogram !! :o [03:04] <ubitux> i missed that one [03:06] <burek> cool [03:06] <burek> i'll use that one for a start [03:06] <burek> thanks (again) :) [07:57] <cone-234> ffmpeg.git 03Kostya Shishkov 07master:5f30c6c8ed1f: Revert "smackaud: clip output samples" [07:57] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:97f7586cd6a1: Merge commit '5f30c6c8ed1f006c1b3a547dfe388f3a4a080e4d' [08:08] <cone-234> ffmpeg.git 03Kostya Shishkov 07master:c3c08baefbf8: smacker: fix off-by-one error in palette expanding code [08:08] <cone-234> ffmpeg.git 03Anton Khirnov 07master:564b7e0c0095: matroskadec: silently skip CodecState element. [08:08] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:91db8ff20304: Merge commit '564b7e0c0095768cd20001b28154d69462be54e7' [08:58] <cone-234> ffmpeg.git 03Anton Khirnov 07master:538bf767373f: avconv: make -aspect work with streamcopy [08:58] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:2ad28840b3fb: Merge commit '538bf767373f176b6487e30181155e74b37cced6' [09:06] <cone-234> ffmpeg.git 03Anton Khirnov 07master:a3fc62e89e1a: configure: make jack depend on pthreads [09:06] <cone-234> ffmpeg.git 03Kostya Shishkov 07master:268bb88544d9: smacker: add a clarification notice about audio decoding [09:06] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:3fc8500483be: Merge remote-tracking branch 'qatar/master' [09:41] <cone-234> ffmpeg.git 03Michael Niedermayer 07release/0.11:ba445be96fa8: avienc: Disallow the first frame to be skiped [09:41] <cone-234> ffmpeg.git 03Michael Niedermayer 07release/1.0:71e32a5ea4dd: avienc: Disallow the first frame to be skiped [09:41] <cone-234> ffmpeg.git 03Michael Niedermayer 07release/1.1:414c6bf09431: avienc: Disallow the first frame to be skiped [09:41] <cone-234> ffmpeg.git 03Michael Niedermayer 07release/1.2:735deda2cfec: avienc: Disallow the first frame to be skiped [09:46] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:cf05fe8a7e58: jpeg2000dwt: remove floats from mixed float/int 9/7 dwt [09:46] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:197e219c927c: jpeg2000dwt: add float based 9/7 dwt [09:57] <cone-234> ffmpeg.git 03Anton Khirnov 07master:c330eba84cf2: lavf: preserve side data when parsing packets. [09:57] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:f6beb3611534: Merge commit 'c330eba84cf2370ad72423aa9e410e20afdc52e2' [10:02] <cone-234> ffmpeg.git 03Anton Khirnov 07master:7b44c9918086: wavpack: remove the subframes codec cap [10:02] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:be72853f8aac: Merge commit '7b44c9918086c6aeacf7ef9f902af38b97f43170' [10:07] <cone-234> ffmpeg.git 03Anton Khirnov 07master:a7ec3a9a4eea: wavpack: drop redundant if/else blocks [10:07] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:84051d57f54f: Merge commit 'a7ec3a9a4eea4e374afcd6e05828b103e6fcb601' [10:30] <cone-234> ffmpeg.git 03Carl Eugen Hoyos 07master:c0957f49c7b3: configure: Fix jack auto-detection, broken since 3fc8500 [10:51] <saste> cehoyos, can you reply to my question about system level libs? [10:57] <cehoyos> saste: I am not sure what to reply, there is a FFmpeg-internal definition of "system library" that has been used forever (?) and I do not remember a user or packager complaining. [10:57] <saste> cehoyos, and what this definition is? [10:57] <cehoyos> I don't think anybody ever suggested that a codec library could be a system library, but imo, it is not unreasonable to count iconv, zlib, bzlib, asound, xv, sdl and pthreads to system libraries [10:58] <cehoyos> And libva [10:58] <cehoyos> That is what we currently do. [10:58] <michaelni> libc [10:58] <saste> do we mark system libraries as such in configure? [10:58] <saste> or is it "implicit"? [10:59] <cehoyos> We mark them in ./configure --help [10:59] <saste> libjack is a system library? [10:59] <cehoyos> But I honestly don't think this discussion leads anywhere: You already agreed that asound and xv are the same beasts, so I think we should agree to not treat them differently. [11:00] <cehoyos> libjack is certainly particularly evil, yes [11:07] <cone-234> ffmpeg.git 03Anton Khirnov 07master:528daa399018: wavpack: switch to planar output [11:07] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:a803ca54e5b8: Merge commit '528daa399018af74d52192eb1861d2b59d256111' [11:07] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:107e9e44329a: wavpack: Properly check channel index, dont write out of arrays [11:12] <cone-234> ffmpeg.git 03Anton Khirnov 07master:474fb0c9d9c6: wavpack: return 0 instead of samples count from decoding functions [11:12] <cone-234> ffmpeg.git 03Anton Khirnov 07master:0f3a0b24dd6f: wavpack: add an error message to a failure. [11:12] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:8d4e934df94e: Merge commit '0f3a0b24dd6f595feaa4526e52ffa7d05c3d7840' [11:28] <cone-234> ffmpeg.git 03Anton Khirnov 07master:1d177200ce1f: wavpack: remove a useless check. [11:28] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:2ba0953ee4bd: Merge commit '1d177200ce1ffd8902729e3a4d647b04fc4b35f8' [11:34] <durandal_1707> it makes me happy that ugly wavpack issue(s) will be finally resolved [11:34] <cone-234> ffmpeg.git 03Anton Khirnov 07master:d1bec4be2207: wavpack: return an error on 0-sized blocks [11:34] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:f5dfc3515780: Merge commit 'd1bec4be2207738669874e02605e879ea9620974' [11:41] <cone-234> ffmpeg.git 03Anton Khirnov 07master:35596bd4f4d3: wavpack: remove a useless parameter from wavpack_decode_block(). [11:41] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:288e3fdfd3ab: Merge commit '35596bd4f4d3c39eed838cf3a2613817d62dda99' [11:47] <cone-234> ffmpeg.git 03Anton Khirnov 07master:72dee5f40b23: wavpack: don't set sample format in init. [11:47] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:aa6cf4c1b050: Merge commit '72dee5f40b236c5eb9e68df39df80c68392269e3' [11:57] <cone-234> ffmpeg.git 03Anton Khirnov 07master:5074f4545c43: wavpack demuxer: export full wavpack blocks. [11:57] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:8012f93ed8af: Merge commit '5074f4545c439420daebe4c2f28ed216440b6ec7' [12:01] <durandal_1707> ugh their apetag writer sucks bigtime, as it requires seekable stream [12:04] <cone-234> ffmpeg.git 03Anton Khirnov 07master:9b6f47c44807: matroskadec: export full wavpack blocks. [12:04] <cone-234> ffmpeg.git 03Kostya Shishkov 07master:851bc1d6be0c: smacker: assign video frame PTS [12:04] <cone-234> ffmpeg.git 03Michael Niedermayer 07master:3ed3aaab32f5: Merge remote-tracking branch 'qatar/master' [12:49] <cone-234> ffmpeg.git 03Paul B Mahol 07master:269fc8e04906: lavf/wvenc: remove hacks since wavpack demuxer & decoder hacks have been fixed [12:55] <cone-234> ffmpeg.git 03Dave Yeo 07master:e615a777990f: os2threads: move from lavc to compat/ [12:55] <cone-234> ffmpeg.git 03Dave Yeo 07master:35e683e457b6: avcodec/vp8: Cosmetics, maintain alphabetical order in threading headers [13:54] <cone-234> ffmpeg.git 03Paul B Mahol 07master:ced0307ea94b: tta: make probing more robust [14:18] <saste> mcdeint gives the same output of mp=mcdeint, but negate,mcdeint and negate,mp=mcdeint differ [14:21] <saste> michaelni, i plan to add an mcdeint test, do I need to add a bitexact option for the snow encoder? [14:21] <durandal_1707> ubitux: tile produces green lines [14:24] <durandal_1707> maybe because after refcounting frames are not bzeroed any more [14:25] <michaelni> saste, i suggest you put the test in a github repo and i merge + test it on a few platforms [14:30] <michaelni> can someone borrow me a recent j2k spec ? [14:32] <durandal_1707> well you can look at libopenjpeg code [14:33] <michaelni> durandal_1707, sure i can but libopenjpeg decodes some j2k files with more artifacts than our decoder [14:34] <JEEB> yeah, looking at another project's code like that in many cases is like looking at GPAC's implementation of something as a reference :D [14:35] <durandal_1707> there is no reference implementation? [14:35] <BuxiNess> michaelni, since 2002 I don't have one [14:35] <michaelni> BuxiNess, mine is FCD from March 2000 [14:37] <Compn> does mike have one ? [14:38] <Compn> BBB : can you get us j2k specs ? [14:38] <BuxiNess> michaelni, in your mbox [14:39] <durandal_1707> Compn: what gives you think that BBB have one? [14:40] <Compn> i think google would have one [14:40] <Compn> chrome probably has j2k decoder [14:41] <michaelni> BuxiNess, thanks alot! [14:41] <durandal_1707> Compn: i seriously doubt that (both statements) [14:41] <Compn> durandal_1707 : i see in firefox bug report that says chrome can open jpeg2000 images [14:42] <durandal_1707> if it uses own lib, i would be first to know that [14:43] <Compn> any strings in chrome binary say it uses openjpeg ? [14:43] <Compn> or chrome license file i guess [14:47] <Compn> maybe chrome went webp instead of j2k [14:48] <Compn> maybe they gave up on it [14:48] <ubitux> durandal_1707: i can't reproduce the problem [14:49] <durandal_1707> ubitux: with what stream you tried? [14:51] <durandal_1707> Compn: well it had support for it via libjasper [14:51] <durandal_1707> but no more [14:51] <ubitux> durandal_1707: matrixbench_mpeg2.mpg [14:52] <ubitux> with 4 seconds instead of 2 [14:53] <ubitux> ah it's not reproducible with png [14:53] <ubitux> but it is with jpeg [14:53] <durandal_1707> i used same args and jpeg output [14:53] <durandal_1707> png does not have subsampling.... [14:59] <ubitux> i've just fixed half of them [15:00] <ubitux> http://pastie.org/7974154 first part. [15:01] <ubitux> here you go [15:01] <ubitux> both fixed [15:02] <ubitux> all the subsampling in drawutils is broken [15:08] <saste> ffplay -i fate-samples/mpeg2/mpeg2_field_encoding.ts -vf "split[a][b]; [a]negate,mp=mcdeint[a1]; [b]negate,mp=mcdeint[b1];[a1][b1]blend=all_mode=difference,lutyuv=y='if(val,maxval,0)',showinfo" [15:08] <saste> this gives every time a different output [15:09] <saste> thus i start to suspect that the problem is not related to the port at all [15:09] <ubitux> -vf quantic [15:09] <saste> ? [15:09] <ubitux> bad joke :( [15:10] <funman> -vf quantic,funman to make it better [15:10] <ubitux> :D [15:10] <funman> ;) [15:34] <ubitux> michaelni: so, http://lucy.pkh.me/v/v-dither%3d0.png vs http://lucy.pkh.me/v/v-dither%3d1.png [15:34] <ubitux> i basically took your prng dithering from mandelbrot [15:34] <nevcairiel> while the second looks better, still rather strong banding in that one [15:34] <ubitux> and tried to apply it as a final bias for each pixel position (same value for each comp) [15:35] <ubitux> with dither=1, i get some strange horizontal lines, and banding is still there [15:35] <ubitux> any idea how i could improve that? [15:40] <ubitux> https://github.com/ubitux/FFmpeg/compare/master...vignette#L3R205 [15:40] <ubitux> this is basically what i'm doing [15:40] <ubitux> same for yuv below [15:41] <saste> michaelni: http://gitorious.org/~saste/ffmpeg/sastes-ffmpeg/commits/port-mcdeint-201305... [15:41] <saste> still no idea about the diff thing [15:56] <durandal_1707> saste: how difference looks like? [15:57] <ubitux> oh i'm stupid i didn't reloaded dv for each x [15:57] <ubitux> ok that's better, still some banding though [15:58] <durandal_1707> you actually writing debanding filter? [15:59] <ubitux> durandal_1707: no i'm just adding some dithering to vignette [15:59] <ubitux> to fix the circle banding effect [16:01] <ubitux> so, just updated: http://lucy.pkh.me/v/v-dither%3d0.png vs http://lucy.pkh.me/v/v-dither%3d1.png [16:01] <ubitux> can we consider the dithering good enough? [16:02] <nevcairiel> i can see banding! [16:02] <nevcairiel> :) [16:02] <ubitux> :( [16:04] <durandal_1707> well if width & height is > 512 banding should not be present [16:06] <durandal_1707> assuming that you see no banding with rgbtestsrc=256x256 [16:07] <durandal_1707> ubitux: you use yuv in vignette? [16:07] <ubitux> the above examples are in yuv yes [16:08] <ubitux> at least it seems [16:09] <durandal_1707> well with dithering its another problem [16:09] <durandal_1707> perhaps it does not work as it should .... [16:19] <michaelni> saste, seems the mcdeint tests pass everywhere i try [16:26] <BBB> Compn: no chrome has no j2k decoder [16:26] <BBB> Compn: to my knowledge at least [16:26] <BBB> and no I don't have specs, sorry [16:41] <durandal_1707> there are reference implamentation in java and c.... [17:09] <saste> michaelni, any idea about the diffing mcdeint issue? [17:09] <durandal_1707> saste: why not just push it, it can be fixed later [17:09] <saste> durandal_1707, i'll push if someone ok it [17:10] <saste> maybe michaelni wants to have a look, since he wrote the original filter [17:17] <durandal_1707> saste: so it happens if negate is used prior to mcdeint, and not if negate filter is used on video and than saved to file, and later run with mcdeint? [17:18] <michaelni> saste, no idea why mcdeint differs from mpcodecs one [17:21] <durandal_1707> make sure that extra scale is not inserted somewhere ... [18:11] <durandal_1707> lol, what funny tiff bug [19:02] <ubitux> mmh in practice it seems that simple dither is good enough [19:02] <ubitux> i guess i'll go for it [19:09] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:f0a2c8285a4e: j2kdec: Try to fix 8bps output case [19:09] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:32475f56f38d: j2kdec/jpeg2000dec: partially merge quantization code [19:10] <ubitux> i wonder if i shouldn't make the filter act like a mask to be blended with another filter [19:11] <durandal_1707> that would usually make things slower [19:19] <llogan> https://lists.ubuntu.com/archives/ubuntu-devel-discuss/2013-May/014553.html [19:22] <durandal11707> llogan: already discussed [19:22] Action: llogan goes to log land [19:22] <durandal11707> nothing interesting to read.... [19:23] <llogan> anyone know how LinuxTag went? [19:23] Action: durandal11707 is waiting for same information [19:24] <Daemon404> you know the fork wars made it big now... most posts i see on 3rd party ml and forums ut it right up tehre with netbsd/openbsd, gcc/egc, etc [19:24] <Daemon404> :> [19:28] <cone-342> ffmpeg.git 03Clément BSsch 07master:1f1df07ce561: lavfi/drawutils: attempt to fix subsampling. [19:29] <durandal11707> ubitux: next step add >8 bit support to drawutils [19:29] <ubitux> nope [19:29] <ubitux> first is to make sense out of ff_draw_round_to_sub [19:30] <ubitux> round_dir >= 0 case seems never used [19:30] <ubitux> and subsampling looks wrong [19:30] <ubitux> i think i'll change that in pad filter [19:31] <llogan> saste: do you think users may actually try to use "yadif=1/3" as an option? (RE: mcdeint) [19:44] <ubitux> mateo`: moving the (text?) metadata at the end doesn't break anything in those software? [19:44] <ubitux> typically, isn't this because id3 parsing is aborted while parsing those strings? [19:45] <ubitux> (and those was never reaching the apic parsing) [19:46] <ubitux> -those [19:47] <ubitux> try removing a few string entry maybe [19:48] <mateo`> it does not seems to break anything with those software since the files from beatport for example have apic frames written at the beginning of the id3 metadata [19:49] <mateo`> i can't say more since those application does not complain about anything when openning a file generated by ffmpeg [19:50] <ubitux> what i mean is that those application might be stopping the id3 parsing because of one of the metadata entries [19:50] <mateo`> they might ... [19:50] <ubitux> basically ignoring the remaining entries [19:50] <mateo`> i'm currently looking a bit closer to the id3 muxer [19:50] <ubitux> also, are you sure there is not a kind of terminator marker placed with ff_id3v2_write_metadata() at the end? [19:51] <ubitux> (would that be related to the recent chaptering introduction?) [19:51] <mateo`> afaik nope [19:52] <mateo`> i found this problem when i introduced the support of apic frames for aiff files [19:53] <mateo`> and didn't take the time to solve it [19:53] <mateo`> ;( [19:53] <mateo`> you know iTunes ... [19:56] <durandal11707> there are markers in id3v2 afaik [19:58] <durandal11707> mateo`: i think iTunes can show id3v2 chapters [19:59] <mateo`> durandal11707: I ll test that feature [19:59] <mateo`> durandal11707: can you tell me more about those markers ? [20:00] <durandal11707> mateo`: typo, should be no markeers.... [20:02] <durandal11707> i have mp3 with chapters, and it have APIC in middle of tags [20:04] <mateo`> can you share a sample ? Does it work with iTunes ? [20:06] <durandal11707> mateo`: http://cre.fm/ [20:14] <ubitux> oh chapters of that mp3 displayed with mpv, awesome :) [20:14] <durandal11707> since ages... [20:14] <ubitux> (in the osd on the cover) [20:14] <ubitux> yeah well it's nice to see that in mp3 [20:16] <durandal11707> there is similar thing for ogg, but i yet to find such files [20:51] <durandal11707> mateo`: also you can try remuxing mp3 with ffmpeg and play that back with iTunes: i'm interested to see will it still show chapters [21:11] <mateo`> durandal11707: iTunes does not seem to support ID3 chapters and the remuxed file (mp3) show the same issue as the aiff ones ... the cover do not show up :( [21:28] <durandal11707> mateo`: but cover and chapters show up in original file? [21:33] <mateo`> durandal11707: cover show up, chapters does not appear on the itunes interface [21:34] <mateo`> durandal11707: i don't know if they are supported [22:24] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:f7129ab928b5: j2k/jpeg2000: fix typo in lut_gain index [22:24] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:aa0d11ed0bdf: jpeg2000dec: remove unneeded returns [22:24] <cone-342> ffmpeg.git 03Michael Niedermayer 07master:fe7136ef8202: j2k/jpeg2000: split stepsize in float & int variables [23:04] <Daemon404> michaelni, ping [23:18] <Daemon404> ubitux, also ping 4 u [23:18] <ubitux> yup? [23:19] <Daemon404> is there any plan to make stuff like 'ffprobe -codecs -of json' output the list n json? [23:19] <Daemon404> in* [23:19] <ubitux> dunno [23:20] <ubitux> shouldn't be that hard to write, i suppose [23:20] <ubitux> i don't plan to do it myself though [23:24] <Daemon404> k [23:27] <saste> Daemon404, => ticket [23:27] <Daemon404> hmm [23:27] <Daemon404> looks like it isnt what i want anyway [23:28] <Daemon404> there seems to be no way to pragmatically get what fourccs we support [23:28] <Daemon404> :D [23:29] <durandal11707> add one.... [23:39] <michaelni> Daemon404, pong [23:40] <Daemon404> michaelni, http://pastie.org/pastes/7976728/text?key=nduxpsmcorwurd50zuhisq [23:40] <Daemon404> does this look pushable to you [23:40] <Daemon404> or have i misunderstood your reply [23:44] <michaelni> Daemon404, LGTM [23:44] <Daemon404> ok [23:45] Action: Daemon404 stares at cone-342 [23:45] <cone-342> ffmpeg.git 03Justin Ruggles 07master:e816a8bc26a9: ffmpeg: use actual packet pts when discarding for the copypriorss option [00:00] --- Wed May 29 2013