[Ffmpeg-devel-irc] ffmpeg-devel.log.20191118

[burek]

[00:18:49 CET] <jkqxz> jamrial:  Pong.
[00:38:01 CET] <jamrial> jkqxz: did you look at my cbs av1 patch, regarding the reference frame state for reading and writing?
[00:39:36 CET] <jamrial> i wonder if my approach is ok, or if we should just remove the ref pointer (it's internal, and only the parser accesses it, so it can easily instead use read_ref), or if we should define a given cbs context as exclusively for reading or writing
[00:50:27 CET] <cone-620> ffmpeg 03Andreas Rheinhardt 07master:7c92eaace2b3: avcodec/cbs: Factor out common code for writing units
[00:50:28 CET] <cone-620> ffmpeg 03Andreas Rheinhardt 07master:cda3e8ca04c0: avcodec/cbs: Fix potential overflow
[00:52:07 CET] <jamrial> also, we should upload some av1 samples. cbs_av1 is untested
[00:53:00 CET] <jkqxz> That's true.  Is there an official compliance test suite yet?
[00:53:15 CET] <jkqxz> (Like H.264.1.)
[00:53:36 CET] <jamrial> yes, but it's mostly different frame size or quantizer values used during encoding
[00:53:45 CET] <jamrial> very few testing assorted features
[00:57:07 CET] <jamrial> https://aomedia.googlesource.com/aom/+/master/test/test_vectors.cc that's the list of samples from the official suite
[01:10:11 CET] <jkqxz> Hmm.  Not sure on that read/write state problem.  The h264_redundant_pps BSF keeps separate read/write contexts, but it really is deliberately editing headers in a way which changes the syntax.
[01:14:34 CET] <jamrial> try the samples i linked, av1_metadata will trip during writing because infer() will look at values the ref state stored after parsing the frame during reading
[01:14:46 CET] <jkqxz> It is neat that the H.26[45] metadata filters can use the same context becauase they don't edit anything which changes the read syntax, but it feels like a slightly tricky feature to be relying on.
[01:14:52 CET] <jamrial> i'm all for a clener solution
[01:15:12 CET] <jamrial> *cleaner
[01:21:57 CET] <jkqxz> Your answer is nice, I think.
[01:23:01 CET] <jkqxz> Unless you want to change the BSFs to use separate contexts, which feels more annoying but maybe more correct if you have other tricky things going on.
[01:23:31 CET] <jamrial> they work as is after applying my fix
[01:24:12 CET] <jkqxz> It doesn't look like it affects any other codecs.
[01:24:13 CET] <jkqxz> H.26[45] only infer things from parameter sets, which won't be overwritten multiple times in a single access unit (excepting the actual edit case, as in h264_redundant_pps).
[01:25:16 CET] <jamrial> what about vp9? it uses a similar ref frame state as av1
[01:25:42 CET] <jkqxz> VP9 reference frames only store properties which will be overwritten with the same thing.
[01:26:00 CET] <jamrial> true
[01:27:33 CET] <jkqxz> The values that you write are always the same as the one slot which you've looked at to pick out the reference information, so you'll always get the same answer.  Maybe that's kindof cheating, though.
[03:31:39 CET] <jamrial> jkqxz: uploaded some samples, will send a patch in a moment
[05:32:27 CET] <cone-285> ffmpeg 03Gyan Doshi 07master:d831edc387c1: avdevice/decklink: add option to drop frames till timecode is seen
[15:04:07 CET] <cone-136> ffmpeg 03Gyan Doshi 07master:f897d8c863c5: doc/filters: improve libvmaf section
[18:41:31 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:73b730e3e688: avfilter/vf_xmedian: add support for alpha formats
[18:41:32 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:a50bd3a50e49: avfilter/vf_atadenoise: support alpha formats
[18:41:33 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:410f81f822de: avfilter/vf_lut2: add 12bit depth alpha formats
[18:41:34 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:f490c71553ac: avfilter/vf_remap: add support for 12bit yuva format
[18:41:35 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:97cf49b7fe1d: avfilter/vf_median: add support for 12bit yuva formats
[18:41:36 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:1b26f2702629: avfilter/vf_gblur: add support for 12bit yuva formats
[18:41:37 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:6a38538a099c: avfilter/vf_avgblur: add support for 12bit yuva formats
[18:41:38 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:86f636348cd5: avfilter/vf_w3fdif: add support for more >8 bit alpha formats
[18:41:39 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:3e524a11d93f: avfilter/vf_bm3d: add support for alpha formats
[18:41:40 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:cb844376c3f6: avfilter/vf_fftdnoiz: add support for alpha formats
[18:41:41 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:76ef2ec4715b: avfilter/vf_bm3d: use boolean for ref option
[18:41:42 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:619f5308122b: avfilter/vf_vaguedenoiser: add support for alpha formats
[18:41:43 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:c7abb073312e: avfilter/vf_blend: add 12bit yuva formats
[18:41:44 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:69f5a77dec48: avfilter/vf_scroll: add support for 12bit yuva formats
[18:41:45 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:b565d6389401: avfilter/vf_maskedclamp: add 12bit yuva formats
[18:41:46 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:be8487798ab5: avfilter/vf_maskedmerge: add 12bit yuva formats
[18:41:47 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:37bf7252969b: avfilter/vf_maskedminmax: add 12bit yuva formats
[18:41:48 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:4b554382a00d: avfilter/vf_neighbor: add 12bit yuva formats
[18:41:49 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:ee8fe9c94d48: avfilter/vf_convolution: add 12bit yuva formats
[18:41:50 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:39a4d69d0512: avfilter/vf_midequalizer: add 12bit yuva formats
[18:41:51 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:a960d3311283: avfilter/vf_chromashift: add 12bit yuva formats
[18:41:52 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:927751076666: avfilter/vf_premultiply: add support for 12bit yuva format
[18:41:53 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:4670f8126adc: avfilter/vf_fillborders: add 12bit yuva formats
[18:41:54 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:2f7da8ca36ff: avfilter/vf_limiter: add 12bit yuva formats
[18:41:55 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:9c85e1a091c1: avfilter/vf_amplify: add support for alpha formats
[18:41:56 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:fa00f8008665: avfilter/vf_deflicker: add support for alpha formats
[18:41:57 CET] <cone-136> ffmpeg 03Paul B Mahol 07master:eae292919b67: avfilter/vf_dedot: add 12bit yuva formats
[21:58:52 CET] <BtbN> Anyone happens to know if Windows Desktop Duplication is capable of capturing fullscreen games and the like?
[22:01:36 CET] <JEEB> it should
[22:01:44 CET] <JEEB> I think someone had a patch for FFmpeg's lavd for it
[22:03:17 CET] <JEEB> ah yes, there's someone's WIP thing
[22:03:40 CET] <JEEB> https://github.com/tsaaristo/FFmpeg/commits/mondup-wip-preview
[22:03:49 CET] <JEEB> BtbN: ^ if you're interested in poking at it
[22:03:55 CET] <JEEB> I haven't been able to review it myself yet
[22:05:40 CET] <BtbN> I found this: https://stackoverflow.com/questions/51613903/desktopduplication-api-produces-black-frames-while-certain-applications-are-in-f
[22:06:03 CET] <BtbN> Which basically translates to: It's tricks, since the video format might change mid capture session, and the function which would notice it returns void and silently fails
[22:06:08 CET] <BtbN> *tricky
[22:06:15 CET] <JEEB> ouch
[22:06:36 CET] <BtbN> But solveable
[22:06:50 CET] <JEEB> yea
[22:07:06 CET] <BtbN> I need it for an event, where we want to stream the desktop, and any fullscreen game, really anything on the screen, out via a Decklink SDI card
[22:07:40 CET] <BtbN> So Desktop Duplication -> decklink seems the most logical choice
[23:58:01 CET] <ePirat> BtbN, can't OBS do that?
[23:58:23 CET] <BtbN> OBS is not headless
[23:58:52 CET] <ePirat> libobs kinda is, if you write code to make it do what you want
[23:59:09 CET] <BtbN> I don't think the decklink output stuff is covered by libobs
[23:59:30 CET] <ePirat> it has ffmpeg output so anything that ffmpeg can output to works in OBS too
[23:59:40 CET] <ePirat> in general, at least
[23:59:46 CET] <BtbN> That's also not part of libobs as far as I'm aware, since it's a plugin.
[00:00:00 CET] --- Tue Nov 19 2019