[Ffmpeg-devel-irc] ffmpeg-devel.log.20180821

burek burek021 at gmail.com
Wed Aug 22 03:05:04 EEST 2018

[09:45:44 CEST] <cone-881> ffmpeg 03Steven Liu 07master:dcef44f13625: avfilter/buffersrc.h: fix typo of the comment
[12:11:32 CEST] <durandal11707> i'm seeking AC-4 files
[12:12:53 CEST] <JEEB> the DASH-IF sample set IIRC has some AC-4 samples
[12:15:35 CEST] <durandal11707> JEEB: where is that?
[12:17:19 CEST] <cone-881> ffmpeg 03Paul B Mahol 07master:42a99b76de98: avcodec: add IMM4 decoder
[12:17:20 CEST] <cone-881> ffmpeg 03Paul B Mahol 07master:a39fdc97b73f: doc/general: add some missed codecs
[12:17:21 CEST] <cone-881> ffmpeg 03Paul B Mahol 07master:6972dae5a494: avformat/avidec: fix demuxing of all keyframes of sample 200707170736151.avi
[12:19:09 CEST] <JEEB> durandal11707: http://testassets.dashif.org/
[12:19:28 CEST] <JEEB> http://testassets.dashif.org/#feature/details/586fb3879ae9045678eab593
[12:19:36 CEST] <JEEB> that's the AC-4 test vectors
[12:20:06 CEST] <JEEB> (if that link doesn't work, that's Features->List->Dolby Audio AC-4
[12:20:34 CEST] <JEEB> as far as I know, that's the only samples of AC-4 I've seen
[12:23:17 CEST] <durandal11707> cool, thank you
[12:25:20 CEST] <j-b> IMM5, IMM6 and IMM3 exist too?
[12:27:55 CEST] <durandal11707> j-b: i doubt IMM3 exist, for IMM5 we have one sample, for IMM6 there are no samples
[14:46:47 CEST] <j-b> Did someone do something cool about audio, this year?
[14:54:03 CEST] <kurosu_> Xiph did some CNN-based denoising ? /o\
[15:02:20 CEST] <kierank> durandal11707: really
[15:02:26 CEST] <kierank> you want to implent dolby ac4
[15:02:27 CEST] <kierank> eugh
[15:15:25 CEST] <cone-881> ffmpeg 03Timo Rothenpieler 07master:dc2d56a66b8a: avcodec/nvenc: add option to turn off a53 closed caption embedding
[16:08:00 CEST] <durandal11707> ffmpeg -i http://streaming.dolby.com/ftproot/mitXperts/ts/ts_multi_codec.ts -f null -
[16:08:08 CEST] <durandal11707> ^ this errors out
[17:14:04 CEST] <cone-881> ffmpeg 03Paul B Mahol 07master:24b24e14d647: avcodec/codec_desc: mark Duck TrueMotion 2.0 Real Time as intra only
[17:21:39 CEST] <atomnuker> wasn't ac-4 just a container?
[17:22:21 CEST] <atomnuker> maybe I'm misremembering, I do remember something about them nih-ing an opus-like codec
[17:22:50 CEST] <durandal11707> anybody have ac-4 decoder?
[17:31:21 CEST] <nevcairiel> ac4 is their object based codec isn't it
[17:31:48 CEST] <kierank> Yes
[17:32:01 CEST] <kierank> FFmpeg isn't going to understand objects
[17:48:22 CEST] <atomnuker> all it would have to do is provide side data about the objects, decoders can output a different amount of channels per-frame
[17:49:17 CEST] <atomnuker> though if each object is its own stream and isn't synced with the rest, welp, not much that can be done
[18:53:52 CEST] <durandal11707> need to make FFmpeg brand
[19:00:05 CEST] <thardin> ffmpeg.org doesn't have nearly enough js or background videos
[19:01:04 CEST] <j-b> background video of 250MB!
[19:02:03 CEST] <gnafu> Encoded using Bink!
[19:03:12 CEST] <TD-Linux> atomnuker, ac4 is a codec, besides object-based I think they much improved the quality over ac3
[19:03:33 CEST] <atomnuker> at least you can download it quickly, what if you have, say, 33 megabyte webpage, with 8mb of css, ~16 mb of js, which take 30 seconds (!) to load fully?
[19:03:50 CEST] <atomnuker> now you can experience this hell by upgrading gmail!
[19:04:26 CEST] <atomnuker> I'm serious, its 33 megabytes, and that's with a blank background, and it does take 30 seconds to fully initialize
[19:15:14 CEST] <durandal11707> why you think FFmpeg brand immediately means js and videos?
[19:18:32 CEST] <gnafu> durandal11707: 'Cause that's what "2!.$).' is all about.
[19:25:47 CEST] <durandal11707> nope
[19:34:47 CEST] <gnafu> durandal11707: I know, I'm just making light (as I'm sure others are).  I agree a good brand language can be beneficial.
[19:36:45 CEST] <gnafu> But given ffmpeg's age and ubiquity, does it need an updated brand to achieve anything useful?
[19:45:59 CEST] <gnafu> If someone wanted to work on that, though, why not?  If I was talented in that area, I could see wanting to contribute my talents to a project I love.
[19:47:11 CEST] <gnafu> durandal11707: Do you have ideas, or know someone with that kind of interest?  I'm no designer, but I'd be happy to provide feedback :-).
[19:51:46 CEST] <durandal11707> gnafu: congratulations, i'm your manager, get to work!
[19:53:19 CEST] <gnafu> durandal11707: :-D
[20:58:29 CEST] <atomnuker> freenode finally did something
[21:08:55 CEST] <cone-881> ffmpeg 03Baptiste Coudurier 07master:b8351ea0c335: avformat/audiointerleave: pad last audio frame
[21:14:59 CEST] <durandal11707> atomnuker: what?
[21:15:39 CEST] <cone-881> ffmpeg 03Baptiste Coudurier 07master:955a989341a5: avformat/mxfenc: fix muxing when audio tracks are longer than video track
[21:17:45 CEST] <durandal11707> ubitux: should 1d cube lut format be supported by lut3d filter? or i should do lut1d filter?
[21:20:59 CEST] <durandal11707> Compn: please add ffmpeg imm4 codec to mplayer
[21:26:17 CEST] <ubitux> durandal11707: it doesn't make much sense in lut3d i think
[21:26:46 CEST] <ubitux> the logic is going to be completely different
[21:27:08 CEST] <ubitux> and typically the options of the filter such as interpolation won't apply
[21:27:29 CEST] <ubitux> now maybe in the source file it could make sense, if you want to share the option parsing
[21:27:44 CEST] <ubitux> but up to you, i don't have a strong opinion on that
[21:28:04 CEST] <ubitux> (option parsing -> lut file parsing)
[21:28:48 CEST] <ubitux> or we could move the lut3d file parsing into a decoder or something, to access the lut themselves as image buffer
[21:29:03 CEST] <ubitux> and use video stream as input for lut3d lut1d ..
[21:29:06 CEST] <ubitux> gtg
[21:29:08 CEST] Action: ubitux &
[21:43:37 CEST] <kierank> TD-Linux: well they had 25 years to do it
[21:46:40 CEST] <TD-Linux> kierank, sadly there aren't any useful samples or comparisons to see how much better it is. I only have rumors
[21:57:12 CEST] <kierank> TD-Linux: it's not a codec there to improve compression imo
[21:57:19 CEST] <kierank> it's their to add as much ipr as possible
[21:57:22 CEST] <kierank> into tvs
[21:57:30 CEST] <JEEB> yup
[21:57:54 CEST] <JEEB> if it is better at something that is just the R&D people actually caring about what they're doing, I guess?
[21:58:21 CEST] <TD-Linux> kierank, right, it's regulatory-capture-the-codec
[21:58:24 CEST] <kierank> probably comparable with opus
[21:59:40 CEST] <TD-Linux> they claim transparency at 96kbps with stereo which would put it in line with opus
[22:27:05 CEST] <j-b> 'morning
[22:37:25 CEST] <atomnuker> j-b: its 21:37, or do you have an even more messed up sleep cycle than me?
[22:37:34 CEST] <j-b> it's always morning on IRC
[22:37:43 CEST] <j-b> and I have a messed up sleep cycle, those days
[23:17:09 CEST] <JEEB> &32
[00:00:00 CEST] --- Wed Aug 22 2018

More information about the Ffmpeg-devel-irc mailing list