<jamrial> michaelni: is there are real world user of 2.8 to keep it maintained?
<jamrial> s/are/any
<jamrial> Lynne: cone just gets confused when new tags get pushed
<michaelni> jamrial, https://trac.ffmpeg.org/wiki/Downstreams lists 3 distros that are not EOL using it
thilo has quit [Ping timeout: 264 seconds]
thilo has joined #ffmpeg-devel
thilo has quit [Changing host]
thilo has joined #ffmpeg-devel
navi-desu has quit [Quit: zzz]
<jamrial> michaelni: i don't think any of those support it officially, but rather they have user provided packages
okx has joined #ffmpeg-devel
thilo has quit [Ping timeout: 240 seconds]
thilo has joined #ffmpeg-devel
thilo has quit [Changing host]
thilo has joined #ffmpeg-devel
\\Mr_C\\ has joined #ffmpeg-devel
navi has quit [Quit: WeeChat 4.0.4]
tufei has quit [Remote host closed the connection]
\\Mr_C\\ has quit [Ping timeout: 255 seconds]
cone-870 has quit [Quit: transmission timeout]
lemourin has quit [Quit: The Lounge - https://thelounge.chat]
lemourin has joined #ffmpeg-devel
jamrial has quit []
MisterMinister has quit [Ping timeout: 260 seconds]
cone-281 has joined #ffmpeg-devel
<cone-281> ffmpeg Lynne release/6.1:HEAD: vulkan: return VK_NOT_READY when no queries are available
MisterMinister has joined #ffmpeg-devel
SystemError has quit [Remote host closed the connection]
SystemError has joined #ffmpeg-devel
AbleBacon has quit [Read error: Connection reset by peer]
<Lynne> 🎉
RT|AO has quit [Ping timeout: 264 seconds]
RT|AO has joined #ffmpeg-devel
<Lynne> ...well, that didn't take long, I have to fix vulkan filters
<cone-281> ffmpeg Zhao Zhili master:891f70c6d555: avutil/hwcontext_vulkan: fix memleak when device_create is skipped
<cone-281> ffmpeg Zhao Zhili master:63078b45999d: avutil/hwcontext_vulkan: cuda doesn't belong to valid_sw_formats
<cone-281> ffmpeg Zhao Zhili master:105657540bdc: avutil/hwcontext_vaapi: return ENOSYS for unsupported operation
georgereynolds8 has quit [Ping timeout: 240 seconds]
georgereynolds8 has joined #ffmpeg-devel
<Lynne> that was easy, it was a 2-line fix
MisterMinister has quit [Ping timeout: 240 seconds]
aljazmc has joined #ffmpeg-devel
aljazmc has quit [Remote host closed the connection]
aljazmc has joined #ffmpeg-devel
aljazmc has quit [Remote host closed the connection]
aljazmc has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
goldsoft has quit [Quit: Connection closed]
goldsoft has joined #ffmpeg-devel
<cone-281> ffmpeg Zhao Zhili release/6.1:bf1e5f277384: avutil/hwcontext_vulkan: fix memleak when device_create is skipped
<cone-281> ffmpeg Zhao Zhili release/6.1:2233b51283d5: avutil/hwcontext_vulkan: cuda doesn't belong to valid_sw_formats
<cone-281> ffmpeg Zhao Zhili release/6.1:1e84d9c5da68: avutil/hwcontext_vaapi: return ENOSYS for unsupported operation
<JEEB> oh, so 6.1 actually got branched.
<elenril> courmisch: so much new simd
<elenril> didn't you say the platform is dead?
<JEEB> I guess I'll try to still get the libaribcaption changes in as soon as I get confirmation from author that the commit message changes are fine by him.
<JEEB> did the review/changes on midnight 23-24, but then posting on the mailing list is what I ended up not capable of doing .-.
<elenril> being JEEB is suffering~
<courmisch> elenril: dead in the silly con valley, still strong in Texas and China
<JEEB> so if the original author is OK with the changes, I'll pull in https://github.com/jeeb/ffmpeg/commits/aribcaption_msz_patches_v3 to master & 6.1, since the author of that module asked me not to have it like it is right now in any release
<JEEB> while there are other things that people requested me to check for 6.1, those I haven't done a full review on yet so it's not going to happen. this at least I went through and rephrased commit messages
<elenril> courmisch: i hear silly cone valley is itself dead
___nick___ has joined #ffmpeg-devel
<courmisch> elenril: did the Big One just happen?
<elenril> metaphorically dead
<courmisch> elenril: scientists claim that the Big One is more likely to level LA than the silly con valley
<courmisch> elenril: I'm not clear what metaphorical death means as applied to a geographical area
qeed_ has joined #ffmpeg-devel
qeed has quit [Ping timeout: 245 seconds]
kepstin has quit [Remote host closed the connection]
kepstin has joined #ffmpeg-devel
<courmisch> in PixBlockDSP, are the strides guaranteed to be multiple of 8/16 even in the unaligned cases?
<courmisch> at least checkasm seems to think yes
kurosu has joined #ffmpeg-devel
<elenril> git grep PixBlockDSP|wc -l
<elenril> 0
SystemError has quit [Ping timeout: 256 seconds]
SystemError has joined #ffmpeg-devel
<courmisch> elenril: how about some helping of -i
<elenril> sounds like a workaround for courmisch bugs
<courmisch> it is a bug that it's called PixblockDSP, but that's not mine
<courmisch> I blame you for failing to correct it
<elenril> sure is crappy signal in this train
<elenril> well it's an internal API with few callers
<elenril> you can go over them one by one, check that it's always true, and document it as a requirement
mkver has quit [Ping timeout: 240 seconds]
navi has joined #ffmpeg-devel
___nick___ has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
___nick___ has joined #ffmpeg-devel
___nick___ has quit [Client Quit]
___nick___ has joined #ffmpeg-devel
dellas has joined #ffmpeg-devel
<durandal_1707> average IQ of doom9 users is bellow room temperature
<haasn> in kelvin?
<durandal_1707> nice try
dellas has quit [Remote host closed the connection]
mkver has joined #ffmpeg-devel
cone-281 has quit [Quit: transmission timeout]
<courmisch> durandal_1707: error[E0369]: binary operation `<` cannot be applied to type `IQ`
<courmisch> error[E0308]: mismatched types
<durandal_1707> courmisch: what are the first 6 letters of word association?
<courmisch> associ?
<courmisch> expected `IQ`, found `Temperature`
<haasn> with the split between verbal and non-verbal IQ scores, we should at the very least expect IQ to be a complex number
<durandal_1707> courmisch: wrong answer, correct answer is: associ
<courmisch> haasn: apparently FFmpeg is stuck in the dark ages of computer programming and has not discovered complex number types
<haasn> durandal_1707: wrong again, correct answer is: w, o, r, d, a, s
<durandal_1707> haasn: 16 letters? yes you are correct. I always agree with ******.
Kei_N_ has quit [Read error: Connection reset by peer]
<haasn> ' ' and ',' are not letters
<durandal_1707> i agree
Kei_N has joined #ffmpeg-devel
<haasn> they are Space and OtherPunctuation, respectively
<durandal_1707> i agree
SystemError has quit [Remote host closed the connection]
SystemError has joined #ffmpeg-devel
jamrial has joined #ffmpeg-devel
<haasn> does huffyuv have a spec?
<haasn> also, does anybody have a pdf for the JPEG2000 spec? (ISO/IEC 15444-1:2019, ITU-T Recommendation T.800 (06/19))
dellas has joined #ffmpeg-devel
Kei_N has quit [Read error: Connection reset by peer]
<Lynne> I got a first edition from december 2000
<Traneptora> link but paywall
<haasn> Lynne: probably fine
<haasn> thanks
<Lynne> we used to have a sekrit dropbox filled with specs
<Lynne> not sure what happened to it, dropbox changed so much, and you needed an account
<Lynne> most of it was smpte specs, though
<haasn> hmm
<haasn> too old to have value 18 in table I-10 'Legal EnumCS values'
<haasn> got it, thanks
<Lynne> you found one, or someone sent it?
<BBB> kierank: it's possible the timing is more-or-less an accident and the terminology is adapted from what we discussed at VDD
<jamrial> haasn: 2015 version is freely available
<jamrial> which has value 18, sYCC
<haasn> hmm
<haasn> so should we mark jpeg2000 images as full range?
<haasn> somehow doing that breaks psnr tests horribly
<haasn> but sYCC is full range right
<Lynne> we should
dellas83 has joined #ffmpeg-devel
dellas has quit [Ping timeout: 240 seconds]
<j-b> "Colorspaces are easy"
<durandal_1707> knight chief President on bridge!
<j-b> :)
<j-b> There is always an XKCD for each geeks' problems
<haasn> one downside to signalling the color ranges as a static list inside AVCodec is that it does not work for codecs which support only full-range gray8 but only limited-range yuv
<haasn> I'm not entirely sure such codecs exist
<haasn> maybe snow?
<j-b> Can we vote people out of the community? Asking for a friend.
<haasn> oh, that is easy to solve by just introducing multiple FFCodec
<haasn> like what pnmenc does
<haasn> it has split between mpeg range pgmyuv and jpeg range pgm
<jamrial> michaelni: you forgot the version bump commits post branching
<durandal_1707> j-b: President can become Tyrant
<j-b> durandal_1707: as can LEADER
<durandal_1707> why we still have LEADER
<j-b> NG
HarshK23 has quit [Quit: Connection closed for inactivity]
<haasn> https://github.com/haasn/FFmpeg/commits/grayscale_sanitization zzz pushed also fixes to handle grayscale consistently (support both limited and full range grayscale)
Krowl has joined #ffmpeg-devel
Krowl has quit [Client Quit]
Krowl has joined #ffmpeg-devel
goldsoft has quit [Quit: Connection closed]
<kierank> BBB: Sure, just wanted to put the dates and times out there. Wasn't suggesting anything.
Krowl has quit [Read error: Connection reset by peer]
Krowl has joined #ffmpeg-devel
<michaelni> jamrial, i forgot more, will fix
Krowl has quit [Read error: Connection reset by peer]
HarshK23 has joined #ffmpeg-devel
otoburb has joined #ffmpeg-devel
<Daemon404> Lynne, that mov patch is real hard to follow...
<Daemon404> ... wait this is from 2019?
<Daemon404> and his last reply was 'woll fix'
<Daemon404> will*
<Daemon404> and that was it?
<Daemon404> now i am confused on why you want me to look at it
qeed_ has quit [Quit: qeed_]
qeed has joined #ffmpeg-devel
cone-579 has joined #ffmpeg-devel
<cone-579> ffmpeg Michael Niedermayer master:9d3a7d30c464: Bump versions prior to 6.1
<cone-579> ffmpeg Michael Niedermayer master:0e15b2b828de: doc/APIchanges: Add 6.1 cut point
<cone-579> ffmpeg Michael Niedermayer master:47e784f88161: Bump versions after 6.1
<cone-579> ffmpeg Michael Niedermayer release/6.1:efac4e2c44c5: Bump versions prior to 6.1
<cone-579> ffmpeg Michael Niedermayer release/6.1:4d476677b0db: doc/APIchanges: Add 6.1 cut point
<courmisch> nooooo! how will the world survive without my RISC-V code
<another|> barely
<courmisch> my world domination plan is foiled again!
<courmisch> Next time, Gadget! Next time!
<elenril> looks like skill issue
<durandal_1707> ban low IQ
<Lynne> Daemon404: mainly because it fixes AAC trimming isses in mov
<courmisch> it sounds more like he is saying FFmpeg is good and "readable code" is nonsense
<Lynne> but it also fixes a few other issues
<Lynne> it is hard to follow, but I'd like to know if it makes sense, to override ctts with stts
<elenril> courmisch: and it sounds just wrong
<durandal_1707> elenril: that is screenshot of your code!
<elenril> i certainly did not write that chunk
<courmisch> elenril: he has a point that code supposed to be readable is not always all that readable
<elenril> sure
<courmisch> though I wholeheartedly agree that many parts of FFmpeg are awful
<courmisch> my religion forbids .c files with more 1200 carriage returns
<elenril> but the claim that you can give up on readability is ....
<courmisch> +than
mkver has quit [Ping timeout: 260 seconds]
<courmisch> hmm, err, how exactly is my SBR sum_square code passing checkasm
<courmisch> while doing literally nothing
<elenril> who tests the testers
<BBB> kierank: fair enough
derpydoo has joined #ffmpeg-devel
<another|> elenril: AI
<Daemon404> Lynne, define 'aac trimming issues'
<Daemon404> it would only possibly affect end trimming
<Daemon404> [15:45] <@Lynne> it is hard to follow, but I'd like to know if it makes sense, to override ctts with stts <-- 'override ctts with stts' is funementally a nonsense statement. theyre different things.
<Daemon404> so: yes, overriding ctts with stts is actully wrong
<cone-579> ffmpeg James Almer release/6.1:5c5d3e315e80: Changelog: mark 6.0
<courmisch> elenril: I guess nothing clobbers the float point result register, so doing nothing counts as returning the same value as the reference C function
<jamrial> err, typo
<cone-579> ffmpeg James Almer master:ff3429991ec1: Changelog: mark 6.1
<Lynne> Daemon404: with the decoder fixed, it fixes the pre-padding of mp4 files generated by fdkaac.c
<Lynne> removes around 3000 samples at the start
<Daemon404> that sounds wrong to me
<Daemon404> the fact anything could be fixed by that patch means something elsewhere is wrong
<Daemon404> the correct fix for priming samples is an edit list
<Daemon404> (you *could* use ctts, but nothing except quicktime will handle audio with a ctts box correctly)
<Daemon404> the patch itself looks wrong/misguided...
<JEEB> (not according to Fraunhofer any more, they now want encoders to strip priming samples)
<JEEB> (and consider edit lists a "hack")
<Daemon404> JEEB, well that fixes it at encoding time. s.
<Daemon404> so*
<Daemon404> it becomes a non issue
<durandal_1707> for new files
<Daemon404> durandal_1707, for old files with priming samples that are not signalle, it is impossible to 'fix' them.
<JEEB> Daemon404: yea just noted since I found it funny as it literally was the only way :P
<JEEB> well as you say, ctts technically was possible
<Daemon404> 99% of everything expects an edit list
<Daemon404> more like 99.999
<JEEB> yea
<Lynne> encoders can't really skip priming samples, it's called algorithmic delay for a reason
<Daemon404> Lynne, they can chop them off and not output them and take the hit in error during decode
<Daemon404> banking on the fact humans are stupid at hearing
<Daemon404> (ofc only works if delay is an integer multiple of framesize)
<Lynne> so they take into account the decoder's delay, and cut it off on their end, such that the time is still correct?
<Daemon404> well. """works""".
<Lynne> good luck to them, it's impossible
<Daemon404> yeah if their delay is e.g. 2048 samples, theyll drop the first two 1024 sample lc-aac frames
<Daemon404> just never output them.
<Daemon404> it's totally possible and widely done. just technically not gret.
<JEEB> yea
<Daemon404> to be clear they drop then from *encoding*
<Daemon404> not decoding
<JEEB> yup
<JEEB> during decoding you just decode
<JEEB> anyways, is there a sample for that mp4 demux patch?
<JEEB> to see what it is supposed to fix
<Daemon404> maybe in the trak
<Daemon404> the subtitle thing?
<Daemon404> using stts there for duration is super wrong anyway, in mov/mp4 the duration of the last sample is determined by min(track_dur, movie_dur)
<Lynne> Daemon404: yeah, I guess so, though it gets complicated with HE-AAC, and anything in the apple ecosystem probably with dumb raw adts
<Daemon404> he-aac is a total crapshoot
<Lynne> yyyyyyyyyup
<Daemon404> you cant really drop frames
<Daemon404> itll never line up
<Daemon404> *has* to be an eitl ist
<Daemon404> edit list*
<Lynne> ah, so maybe something's wrong with edit list parsing, either on our end, or fdkaac.c
<Daemon404> is this during encoding or decoding?
<Daemon404> is the encode done by ffmpeg?
<Lynne> decoding, I did say fdkaac.c == command like fdkaac
<Daemon404> are you outputting to mp4 or adts?
<Lynne> mp4, the default
<Daemon404> curious if their cli tool is failing to write an edit list
<Daemon404> aka leaving it totally unsignalled
<Lynne> it writes something, because apple's afconvert-generated files are not changed by the patch
<Daemon404> can i see a fil?
<Daemon404> file even
<Daemon404> i swear to $deity if this is more youtube-edit-list-patch shenanigans...
<Daemon404> thanks
<galad> that files uses the iTunSMPB metadata
<j-b> jamrial: ping
<jamrial> pong
<Daemon404> galad, yep... was just checking if iTunSMPB had proprietary delay data
<galad> yup
<Daemon404> or if the M4A brand implied it
<Daemon404> because there is no els
<Daemon404> elst*
<galad> yet another way to signal the same thing
<Daemon404> ffs apple
<Daemon404> THERE IS A STANDARD
<galad> the funny thing is that when you open that with avfoundation, it converts the iTunSMPB metadata to an edit list
<Daemon404> ...
<Daemon404> imagine my face right now.
<jamrial> Daemon404: Apple decides what is standard. everyone else follows
<Daemon404> Lynne, looks like we need to add iTunSMPB parsing
<Daemon404> trying to find its """spec"""
<durandal_1707> ban apple
<Daemon404> jamrial, true.
<galad> iTunes & iPod teams made some weird things back in days, I wonder why when they had already a solution
<Daemon404> cant find a spec but some old android oss code has a parser
* JEEB stabs his development VM
derpydoo has quit [Ping timeout: 240 seconds]
<Lynne> is that a third way of handling track pre-skip?
<elenril> >Apple decides what is standard. everyone else follows
<elenril> I almost miss the days when microsoft was the company doing this
<Daemon404> hmmm
<galad> Lynne: yes
<Daemon404> hmm
<galad> it was introduced when Apple added gapless playback to iTunes
<Daemon404> we have *something* for it in mov.c
<JEEB> start_pad :3
<Daemon404> hmm
<Daemon404> we *do* set skip_samples properly on Lynne's file
<Daemon404> so it should be trimmed properly
<Daemon404> it's even the right value
<Daemon404> "side_data_type": "Skip Samples",
<Daemon404> "skip_samples": 2048,
<Daemon404> on the first packet
<durandal_1707> ffmpeg.c ignores that
<Daemon404> i thought skip_samples was handled by avcodec?
<Daemon404> Lynne, i have a guess: do you know if our decoding is off by the full 2048 samples, or maybe just 1024?
b50d has joined #ffmpeg-devel
<durandal_1707> iirc avcodec just sets signal
<Lynne> depends on the encoder
<Daemon404> well. on this fil.
<Daemon404> file*
<Daemon404> durandal_1707, are you saying ffmpeg.c ignores skip_samples?
<Daemon404> and avcodec too?
<Daemon404> that cant be right
<Lynne> with fdkaac/most other encoders, it's 2048 samples, only with apple it's actually 2112 samples, with our aac encoder, it's 1024 samples (the algorithmic delay), and when you go over to HE-AAC, like that file, good luck
<Lynne> with fdkaac with that file it's 5057 samples IIRC
<durandal_1707> Daemon404: there are multiple ways in API to skip samples, one of them is 100% ignored by ffmpeg, but mpv use it.
<Lynne> fixing the SBR priming such that it's as the spec requires, and I still get a large gap at the start
<Daemon404> Lynne, ok. i was asking because i noticed the packet skip_samples (with value 2048) is attached to is 1024 samples long, so i thought maybe something was failing to apply skip_samples to the stream, and only to the first packet
<Daemon404> but that would be off by 1024, not 2048.
<Daemon404> so my guess ws wrong.
<Lynne> somehow every single encoder is wrong, and the third least competent encoder is right
<durandal_1707> iirc i poked elenril to fix ffmpeg.c for that one, but he ignored my requests.
<Lynne> I have no explanation why they're using 2048 sample pre-skip, even if they had a lookup
<Lynne> durandal_1707: the side data is never ignored
<elenril> durandal_1707: slander
<Lynne> the decoder-provided avci->skip_samples is ignored if there's side data
<elenril> i challenge you to a duel
<Daemon404> Lynne, well somewhere in the pipeline something is failing to apply_skip samples properly
<Daemon404> the delay is correctly set at demux
<Daemon404> apply skip_samples(
<Daemon404> *
<Lynne> why does only that specific file get better with the patch?
<Daemon404> i dont know, pure luck
<Daemon404> because what it changes is super unrelated
<durandal_1707> look at my fix for opus skip samples
<Daemon404> link?
<cone-579> ffmpeg TADANO Tokumei master:82faba8a6ce8: lavc/libaribcaption: switch all `bool` context variables to `int`
<cone-579> ffmpeg TADANO Tokumei master:21bfadd9b4a2: lavc/libaribcaption: add MSZ character related options
<cone-579> ffmpeg TADANO Tokumei master:a824c6f2f627: lavc/libaribcaption: rename `replace_fullwidth_ascii` to `replace_msz_ascii`
<j-b> JEEB: thx for pushing those.
<JEEB> j-b: I will attempt to put those into release/6.1 as well since the library author requested the first release to include the wrapper to have these changes
<JEEB> I did the review in midnight 23-24 but then crap happened so I never got to posting what I just did earlier
<j-b> JEEB: the issue is that this changes the libaripcaption requirement, but I think it would be fair
<JEEB> yea, he requested it to be bumped. 0.1.0 did not build to begin with
<JEEB> you had to go X commits after 0.1.0 to begin with
<j-b> yeah, but the MSZ commit requires 1.0.x no?
<JEEB> yup
<JEEB> anyways the requirement bump was a request from the author since otherwise there was messy ifdefs
<JEEB> and since the wrapper was not yet in any release
<Daemon404> oh good there is debug code for skip_samples
<Daemon404> might help
<JEEB> I also checked the common linux distros and aribcaption doesn't seem to be packaged yet
* Daemon404 rebuilds as debug
<JEEB> so if they package, it's going to be latest
<JEEB> homebrew on mac has already updated, I think?
<JEEB> but at the end of the day, the author of the aribcaption library requested the requirement to get bumped
<JEEB> (since older versions had issues)
<Daemon404> i have a suspicion again
* Daemon404 stares at decode.c
<JEEB> :D
<Daemon404> (though i cant find durandal_1707's opus fix)
<jamrial> Daemon404: Lynne was going to touch that skip_samples stuff afaik
<JEEB> so in cherry-picks, do we also take in the micro bump and just match it against the minor of the release branch?
<Daemon404> i mean thats who i was talking to :P
<jamrial> right :
<Daemon404> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x559b6d6e0580] demuxer injecting skip 2048 / discard 0
<Daemon404> [aac @ 0x559b6d6e1d80] skip 2048 / discard 0 samples due to side data
<Daemon404> this seems suspect
<Daemon404> [aac @ 0x559b6d6e1d80] skip whole frame, skip left: 0
<JEEB> so in this case since both master and release/6.1 are at micro 100 what was xxx.32.101 on master would be xxx.31.101 on release branch?
<JEEB> or do I just leave the micro bump out?
<Daemon404> because the frame is 1024 sampls
<Daemon404> how can there be 0 skp left?
<Daemon404> OH
<Daemon404> duh
dellas83 has quit [Remote host closed the connection]
<Daemon404> the skip is 2048 SAMPLES, and the frame is 1024 ... in timbase 24000
<JEEB> \o/
<JEEB> conversion from one to other?
<Daemon404> so makes sense to skip exactly one avframe
<JEEB> ah
<JEEB> right
<Daemon404> so the first PTS ends up being non-0
<Daemon404> will ffmpeg.c filter stuff inject silence?
<Daemon404> (which would add the delay right back lol)
<cone-579> ffmpeg TADANO Tokumei release/6.1:48afb4354902: lavc/libaribcaption: switch all `bool` context variables to `int`
<cone-579> ffmpeg TADANO Tokumei release/6.1:8ccd1593a4b8: lavc/libaribcaption: add MSZ character related options
<cone-579> ffmpeg TADANO Tokumei release/6.1:1cff6e41bf2b: lavc/libaribcaption: rename `replace_fullwidth_ascii` to `replace_msz_ascii`
<Daemon404> actuallly maybe output is correct...
<JEEB> there. that is all from me since this was my mistake of doing review between 23-24 but then not having the mental to send the final OK on the commit message rewrite + the library author had requested the first release including that wrapper to include those changes.
<Daemon404> Lynne, i you perhaps was your input to fdk a wav?
MrZeus has joined #ffmpeg-devel
<Daemon404> (and if so from where)
<jamrial> JEEB: things that bump version should not go to release branches
<JEEB> even micro?
<Daemon404> i am almos wonderng if there's priming samples left over from a prevous encode/decode cycle
<Daemon404> before it even got encoded with fdk
<Daemon404> hmmm no, maybe not
MrZeus_ has joined #ffmpeg-devel
<JEEB> minor of course not
<Daemon404> it is very odd, the avcodec output (in avframes) looks correct
<jamrial> JEEB: eh, i guess micro is not a problem, as long as the addition is also in master
<Daemon404> but ffmpeg.c output is *not*
<durandal_1707> git misbehaves baddly here, can track libavcodec/opus_parse.c changes
<Daemon404> this sounds cursed
<JEEB> jamrial: yea, it is. and also got requested from the library author to not have the first release including this wrapper without these changes
MrZeus__ has joined #ffmpeg-devel
<durandal_1707> opus_parse set delay and skip samples
<Daemon404> skip_Samples is being set here already correctly
MrZeus has quit [Ping timeout: 240 seconds]
<JEEB> jamrial: but in general, for micro bumps - has the standard so far been that such fixups' bumps don't end up in cherry-picks?
<durandal_1707> time to update buggy git
<durandal_1707> Daemon404: note that ffmpeg.c ignores initial_padding for sure
<durandal_1707> huh, no update for buggy git?
MrZeus_ has quit [Ping timeout: 240 seconds]
<durandal_1707> everything is going downhill
<Daemon404> durandal_1707, what i suspect is happening is the aresample flter is inserting silence since first avframe has non-zero pts due to skip_samples being applied in decode.c
<jamrial> JEEB: i don't recall cherry-picks that introduced avoptions
<Daemon404> decode.c will drop output but not shift timestamps
jamrial has left #ffmpeg-devel [#ffmpeg-devel]
jamrial has joined #ffmpeg-devel
<durandal_1707> Daemon404: sample ?
<Daemon404> [16:13] <@Lynne> https://0x0.st/Hyj2.m4a
<Daemon404> ^ durandal_1707
<durandal_1707> aresample should not insert silence by default, that is meant as additional feature to be enabled by option
<Daemon404> i see
<Daemon404> well output has exactly 2048 samples of silence at the start
<Lynne> Daemon404: raw
<JEEB> jamrial: yea I don't expect to be doing that again any time soon. just got heavily requested to not have the wrapper without those changes in the first release the wrapper would be in
<Lynne> not wav
<jamrial> JEEB: that's ok
<Daemon404> Lynne, from a CD?
<durandal_1707> start_pts=1024
<Daemon404> i.e. never been encoded before
<Daemon404> durandal_1707, which is correct, because the first decode frame is droppe
<Daemon404> due to skip_samples
<JEEB> jamrial: but yea, normal fixups don't seem to get micro bumps around here
<Lynne> Daemon404: no, some random .flac file I found, I trimmed it, and outputted it to 48khz 2channel raw
<JEEB> so I guess the question of whether to back-port micro bumps is not that big of a thing
<Daemon404> Lynne, ok
<Daemon404> i suspect ffmpeg.c is to blame atm
<jamrial> JEEB: yeah, we don't normally bump micro for fixes, only for changes that are exposed in some form to the api user, as is the case of avoptions
<JEEB> right'o
<Lynne> Daemon404: for the part that isn't the additional SBR decoder delay, right?
<Lynne> let me send you my patch which fixes it
<Daemon404> Lynne, specifically due to priming, not SBR
<Lynne> ah, really?
<Lynne> if I manually skip via an option I added, it's correct, so it could be the side data, perhaps somewhere in decode.c
<Daemon404> ohhhh fuck
<Daemon404> i just realized this is he
* Daemon404 deskface
<Daemon404> so maybe ONLY priming skip is applied
<Daemon404> when it should be sbr+priming?
<Lynne> I fixed that part, I added a += sigh in decode.c so both are taken into account
<Lynne> but skipping the decoder-specified number of samples as per the spec still leaves a gap, even with both the side data and this
<Daemon404> 'gap' here meaning extra silence at the start?
<Lynne> yup
<Daemon404> or is this specifically when measurin *between* tracks being playe
<Daemon404> (i.e. could be end padding)
<Lynne> no, at the start
<Daemon404> ok
b50d has quit [Remote host closed the connection]
MrZeus_ has joined #ffmpeg-devel
<Daemon404> well ffmpeg.c output looks correct for the amount of skip_samples
<Daemon404> Lynne, so this only happens with HE right?
<Lynne> err, I just tested it, and it works now
<Daemon404> if adding sbr to the skip works, i guess that is the right fix
<Daemon404> yeah
<Daemon404> i thought i was going crazy lol
<Lynne> no, it's just me, I'm having to deal with it
<Lynne> I can't believe we don't skip any samples during seeking
MrZeus__ has quit [Ping timeout: 248 seconds]
<Lynne> it's very sadistic to seek in he-aac and hear +12dB white noise burst
<Daemon404> gross
<elenril> serves you right for seeking
<j-b> Seeking?
<elenril> only uncultured plebeians seek
<Lynne> seek button is mandatory nowadays, especially in youtube content
<Daemon404> imagine telling christopher nolan you seek in his movies
<Lynne> it's not an IETF RFC, it's something higher, an internet standard
<elenril> >implying I watch youtube through a browser
<Lynne> exactly
<Daemon404> yeh elenril uses ipad only for YT
<elenril> the virgin seek button
<elenril> the chad playback speed
* elenril stabs Daemon404
<durandal_1707> elenril is Master Chad
<Lynne> why would anyone want a slider to make voices sound like chipmunks?
<Daemon404> fetishes
<JEEB> the voices get adjusted
<Lynne> there are already enough nightmare fuel movies out there which do that
<JEEB> I watch most of my youtube at 1.75x
<durandal_1707> Lynne: seek in mpv works that way anyway, even for flac
<JEEB> or well, pitch adjustment
<elenril> Lynne: like you you didn't watch chip&dale as a kid
<Lynne> flac does not have any decoder delay
<durandal_1707> git log libavcodec/opus_parse.c
<durandal_1707> why i can not follow full history of this?
<durandal_1707> all other commits are unreeachable
<Daemon404> did it get renamed
<durandal_1707> yes
<durandal_1707> --follow does not help
<Daemon404> git log --follow
<Daemon404> o
<Daemon404> rip
<durandal_1707> can you also reproduce that?
<Lynne> checkout the previous commit and go from there is my way
<Daemon404> the refactor must have been too hardcore
<Daemon404> so git cant figure out where it came from
<Daemon404> do what Lynne said
<elenril> Daemon404: it clearly can, git show does see the move
<Daemon404> git show what?
<elenril> git show the commit that created the file
<j-b> git blame -C -M
<Daemon404> show doesnt show a move
<Daemon404> just a new file
<elenril> do you even --color-moved
<Daemon404> elenril, still dont see it.
<Daemon404> just new file. even with that flag.
<durandal_1707> Waiting for data... (^X or interrupt to abort)
<Daemon404> elenril, i even redid the commit:
<Daemon404> 7 files changed, 595 insertions(+), 529 deletions(-)
<Daemon404> create mode 100644 libavcodec/opus_parse.c
<Daemon404> create mode 100644 libavcodec/opus_parse.h
<Daemon404> it does *not* recognize it as a move
<Daemon404> (you would see x -> y (n%))
<jamrial> Daemon404: -M option from format-patch doesn't do anything?
<Daemon404> i didnt use format-patch.
<Daemon404> that also doesnt help git log, which is what durandal_1707 is using.
dellas has joined #ffmpeg-devel
_whitelogger has joined #ffmpeg-devel
goldsoft has joined #ffmpeg-devel
goldsoft has quit [Client Quit]
goldsoft has joined #ffmpeg-devel
psykose has quit [Remote host closed the connection]
psykose has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
dellas has quit [Remote host closed the connection]
dellas has joined #ffmpeg-devel
<ePirat> good lord ffmpeg ML is annoying today
<durandal_1707> lol
<another|> So... like any other day?
<Lynne> it's not as bad today, but it's bad
<Daemon404> it's been extra trashfire lately
<Daemon404> people keep adding fuel
<Lynne> yeah, I blame the financial thread and "what did you just say to me" ping pong 6.1 release thread
<durandal_1707> sorry state
<ePirat> also Nicolas' mails break my brain…
<Daemon404> im sad nobody used the navy seal copypasta
b50d has joined #ffmpeg-devel
AbleBacon has joined #ffmpeg-devel
<elenril> clearly you must save the day
<goldsoft> I want to confirm whether ffmpeg does not support uploading via WebDAV.
<goldsoft> The IPC stream acquisition is normal, and the file upload test with curl is also normal, but it is unable to upload with FFMPEG
<durandal_1707> goldsoft: stop spamming here!
<goldsoft> ok
b50d has quit [Remote host closed the connection]
<another|> I mostly don't read these gigantic threads. Especially those that split off into new threads every now and then
<Son_Goku> Lynne: I just want ffmpeg 6.1 already...
<another|> the 6.1 mails are in at least 2 threads with 120+ mails in my inbox
<Son_Goku> I think we managed to not break ABI this release so I can upgrade ffmpeg in Fedora...
<Lynne> branch has been made, you'll get it soon
<elenril> don't worry, the next one will be broken
<Son_Goku> worst comes to worst, 6.1 will be F40+ only
<Son_Goku> but I'm hoping I can bring it back to F38 and F39
<another|> hmm.. thunderbird marked NG mail as spam
<another|> ... but it also marked one from courmisch
<Lynne> elenril: all our releases are broken, except 2.8, which has had 8 years to stop being broken
<Lynne> we should release versions like wine is released
<elenril> nah, all releases without fully threaded ffmpeg CLI are broken
<durandal_1707> including one with threaded ffmpeg CLI
<elenril> my threaded branch is a work of art
<another|> Lynne: wine as in drink or software?
<Lynne> well, yeah, when that gets fixed, there's going to be something else that's broken, like ffmpeg .ppt support
<elenril> it's made to be admired, not used
<JEEB> ffmpeg doesn't make kvass for me
<durandal_1707> sonic is still broken
<durandal_1707> sws still broken
<durandal_1707> swr still broken
<durandal_1707> avdevice still broken
<Lynne> snow is broken because I don't see snow as the output
<durandal_1707> postproc is broken
<durandal_1707> subtitles broken too
<elenril> avfilter is the brokenest of them all
<Lynne> that all still leaves avutil, which is less than broken, just occasionally non-functional
<durandal_1707> elenril: what is broken?
<elenril> it's just hwcontext that qualifies lavu for non-broken status
<elenril> hwcontext is the best
<elenril> durandal_1707: what isn't
<JEEB> anyways, hopefully I'll finish my avctx side data stuff for 7.0
<JEEB> since jamrial got ideas from me and got his stuff in first :D
<ePirat> Son_Goku, are you the ffmpeg fedora package maintainer?
<Son_Goku> yes
<another|> Lynne: `ffmpeg -f rawvideo -video_size hd720 -i /dev/urandom -t 5 -c:v snow foo.nut`
<ePirat> Son_Goku, thanks for your work!
<Son_Goku> you're welcome
<Lynne> fedora has an ffmpeg package?
<JEEB> yes
<JEEB> even mpv, I was surprised
<JEEB> and libplacebo
<Lynne> how? we're all so full of patented algorithms despite never reading any
<JEEB> of course for most formats you need freeworld because of US software patents
<durandal_1707> elenril: the threading work is full of hacks and does not really work unlike avfilter
<JEEB> so "totally unrelated even though maintained by same people" RPMFusion -freeworld packages exist
<Lynne> I remember it was a requirement that all patented code must be removed
<JEEB> at some point that changed apparnetly
<JEEB> and build config only became a thing?
<Daemon404> the only way to be sure no patented code sneaked in is rm -rf /
<durandal_1707> everyhing is patented
<another|> >inb4: patent on unlinking data from a filing system
<microchip_> durandal_1707: even you! I patented you yesterday :D
<Daemon404> method to remove nodes from a tree
<JEEB> method and apparatus
<Daemon404> oh yes
<j-b> I would say that the only way to be sure that there is no patent, is that the patent existed and is now expired.
<j-b> Like MPEG-2 video or AC-3
<durandal_1707> they repatent stuff
<JEEB> j-b: yea, knowing that you are working with a patent that already existed and expired means that you have something to point at
<Lynne> you can necromancer dying patents, usually companies do this once or so
<Son_Goku> JEEB: actually the rpmfusion package is maintained by different people
<Son_Goku> Lynne: it happens a lot less than you'd think, thank god, because the USPTO charges an absurd amount of money and it's hard to do without getting the patent invalidated
<Son_Goku> Sonos recently got screwed that way
<Son_Goku> Daemon404: the packaging for ffmpeg in Fedora is *special*
<Son_Goku> I maintain an allowlist of codec files and generate a tarball based on that
<Son_Goku> and then further maintain a list of allowed to enable stuff
<durandal_1707> see, 99% stuff is disabled/
<j-b> Are you allowed to call FFmpeg if all is disabled?
<elenril> durandal_1707: threading work passes fate in its entirety, except for sub2video
<Son_Goku> durandal_1707: it's not 99% disabled actually
<another|> --enable-calling-it-ffmpeg
<Son_Goku> j-b: why wouldn't we? it's no different than someone else doing an ffmpeg build where only some codecs are enabled
<Son_Goku> it's part of the build scripts to allow that
<j-b> Son_Goku: just asking. But for VLC, this is not allowed.
<Son_Goku> it's pretty common for people to ship builds of ffmpeg where not every single codec option is enabled
b50d has joined #ffmpeg-devel
<JEEB> yea, ubuntu after all used to default to it
<JEEB> then apparently they stopped doing that at some point
<Son_Goku> they definitely still do
<Son_Goku> since they inherit it from debian
<Son_Goku> which still does
<Son_Goku> the difference between libavcodec and libavcodec-extra in debian/ubuntu
<JEEB> yea
<JEEB> but it's no longer "no H.264" etc
<j-b> depends how much you disable
<j-b> Also, it's disable, not enable-just-a-few
<Son_Goku> JEEB: ubuntu isn't separately maintaining it from debian, and I'm surprised debian is doing that since their own policy forbids it
<Son_Goku> but then again, it's actually quite hard to enforce debian policy on debian maintainers
<JEEB> Son_Goku: I think whatever changed happened around ~10 years ago
* Son_Goku shrugs
<Son_Goku> I'm working with what I can do
<JEEB> anyways since I see most users understanding that it's a distro problem and not an FFmpeg problem I don't get too annoyed with this stuff since debian/ubuntu used to have this as well in the olden days. of course with debian/ubuntu the easy part was that it was just another package in already enabled repos
<Son_Goku> the list of codecs we have not-enabled is pretty small
<Son_Goku> and our arrangement with Cisco for OpenH264 also closes the gap pretty well
<Son_Goku> so our ffmpeg can play H264 through that codec library
<JEEB> yea, cisco did a good thing there
<JEEB> since they're already paying max
<Son_Goku> I wish that the ffmpeg-libopenh264 support wasn't quite so barebones, but I just don't understand how this stuff works to fix it
<Son_Goku> e.g. no support for SEI stuff
<Son_Goku> but it's better than nothing
<Son_Goku> and works in most cases
<JEEB> yea that's the classic thing where the built-in functionalities in the framework are not utilized due to it being a 3rd party wrapper
<JEEB> you could in theory start mapping all that stuff, but I'd bet most people just don't see the benefit
<JEEB> and/or only utilize openh264 for encoding
<JEEB> since most people utilize the internal h.264 decoder which naturally has everything wired up already
<Son_Goku> yeah
<Son_Goku> we still have that disabled unfortunately
<Son_Goku> I'd like to enable it at some point, and I hope I can soon
<durandal_1707> disable FLAC too
<Daemon404> there is probably a way to expose stuff
<durandal_1707> its full of patents
<Son_Goku> :/
<Son_Goku> flac has been allowed in Fedora for a very long time
<Son_Goku> durandal_1707: it's not a question of patents or not, it's a question of a useful patent grant
<Son_Goku> that's why the VPx and AV1 codecs are in Fedora
<kierank> does fedora disable aac?
<Son_Goku> nope
<kierank> he-aac is still patented
<Son_Goku> well, we have a stripped fdk-aac
<Lynne> isn't mpeg-4 aac still under patent too?
<Son_Goku> sort of?
<kierank> Son_Goku: yeah but decoder?
<kierank> he-aacv2 clearly patented
<j-b> clearly
<Daemon404> people serving he-aac to people get what they deserve
<JEEB> Lynne: I think the patents affect features? (and thus mpeg-4 audio types)
<Son_Goku> yes
<j-b> yes
<kierank> so how is it ok in fedora to ship?
<Son_Goku> we don't have those things
<j-b> You do.
<kierank> you have a patched ffmpeg with he-aac removed?
<kierank> (not fdk, ffmpeg)
<JEEB> they probably disable avcodec decoder
<JEEB> and only enable fdk-aac
<Son_Goku> yup
<JEEB> which is then patched
<Son_Goku> yup
<j-b> So, insane of supporting the FFmpeg decoder, you are using a 3rd party library who has very doubtful license
<kierank> j-b: no
<kierank> they are shipping aac
<j-b> instead*
<kierank> first one on the list
<kierank> shipping dolby e too lol
<kierank> definitely patented
<j-b> but no HE-AAC?
<kierank> in ffmpeg you get the whole thing
<Daemon404> dolby is a kind company who would never sue anyone
<kierank> if you ask for aac
<j-b> since when is AAC no patented?
<j-b> if it is shipped?
<kierank> they have no patch to remove he-aac(v2) from ffmpeg
<Son_Goku> *sighs*
<Son_Goku> fuck
<j-b> I would be surprised
<Son_Goku> well, I'm auditing everything by hand
<Son_Goku> so I guess I need to go and clean stuff out partially then
<j-b> Because SBR is patented.
<kierank> with your lawyer sitting next to you
<j-b> So, not only is HE-AACv2 not free, HE-AAC isn't either.
<Son_Goku> you seem to make a ton of assumptions here
<Son_Goku> I'm just a guy in Fedora doing stuff
<Son_Goku> I have to do crazy back and forth to get things in good shape
<kierank> who wrote that fdk patch to remove he-aac?
<Son_Goku> that was wtay at Red Hat
<kierank> 19:50:32 <j-b> So, insane of supporting the FFmpeg decoder, you are using a 3rd party library who has very doubtful license
<kierank> if they have removed he-aac(v2) then the licence is fine
<j-b> ahaha
<kierank> if they can prove the code has not patents attached
<j-b> I told wtay a long time ago that his patch was wrong
<j-b> he did not care
<j-b> because SBR is patented
<j-b> so he should remove HE-AAC too
<Son_Goku> he did
<Son_Goku> we don't have SBR in fdk-aac
<j-b> but he only removed HE-AACv2
<Son_Goku> no, both are gone
<j-b> ok then.
<kierank> so why all the fuss about fdk-aac, yet aac in ffmpeg gets shipped
cone-579 has quit [Quit: transmission timeout]
<Son_Goku> kierank: because I did not know there was HE-AAC code in ffmpeg, the docs stated only AAC-LC
<Son_Goku> which is allowed
<j-b> aacsbr.c
<kierank> lol
<JEEB> Son_Goku: yea the docs quite often be old and misleading regarding what media formats are implemented
<Son_Goku> as I'm learning right now
<j-b> PNS is not patented?
<Son_Goku> expired
<JEEB> different types of HE-AAC have been implemented for quite a few years
b50d has quit [Remote host closed the connection]
<Son_Goku> though actually I need to double-check on SBR
dellas has quit [Remote host closed the connection]
dellas has joined #ffmpeg-devel
cone-241 has joined #ffmpeg-devel
<cone-241> ffmpeg Rémi Denis-Courmont master:04b49fb3c5ad: lavu/riscv: fix typo
<Son_Goku> anyway, at this point I'm just waiting for VLC 4
<Son_Goku> which will hopefully release sometime soon
<elenril> lol
jamrial has quit []
jamrial has joined #ffmpeg-devel
<durandal_1707> elenril: have you incorporated amix fix into your set?
<jamrial> <@JEEB> since jamrial got ideas from me and got his stuff in first :D <-- i did not end up using a side data set struct :p
<JEEB> sure
<elenril> durandal_1707: yes
<JEEB> but side data in avctx
<Son_Goku> so when VLC 4 j-b :P
<BtbN> haasn: did you ever end up getting a GPU from Nvidia btw.? I still have my 3070 here which I'd gladly send you otherwise.
aljazmc has quit [Quit: Leaving]
___nick___ has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
<Lynne> I thought you sent yours off?
___nick___ has joined #ffmpeg-devel
___nick___ has quit [Client Quit]
___nick___ has joined #ffmpeg-devel
<Daemon404> we're gonna need a bigger dumpster
<Lynne> stop responding and ignore, you've all been on the internet long enough?
<courmisch> can't even do some autoderision, as it's insinuating being insulted
<Daemon404> i dont respond on the ML to it
<Daemon404> because i dont want to hate life.
<Daemon404> but if noone responds then ffmpeg goes to shit
<Daemon404> so.
<durandal_1707> both ways - loose.
<Daemon404> indeed
<Daemon404> nihilism ftw
<courmisch> some people are going to have some very real life problems
<courmisch> if their employment problem is not resolved
<elenril> sounds kinda tautological
<Daemon404> problems wont just be for people employed to work on ffmpeg
<jamrial> i don't understand what is happening on that spi thread
<courmisch> I really want to answer "please refrain from insinuating that I am insinuating that you are insulting me"
<courmisch> but I won't
<Daemon404> jamrial, crack cocaine
<elenril> thank you for helping us help you help us all
<courmisch> jamrial: El Presidente threatened to quit and go make money on non-multimedia startups...
<courmisch> jamrial: ...and now people are panicking that they need to make a living
<Daemon404> ironcally making the former more likely
<Daemon404> and killing it more
<j-b> jamrial: and the accusation against Kieran?
<Daemon404> he is projecting
<Daemon404> just wait for it.
<BBB> I do think nicolas' statements are hurtful and false
___nick___ has quit [Ping timeout: 255 seconds]
<BBB> and I'm thankful that Michael apologized for hit statements to kieran. that was necessary
<BBB> nicolas could learn from that
<Daemon404> thilo hasnt
<BBB> indeed...
<j-b> People who don't apologise should risk sanctions.
<kierank> I have sent two emails to the CC
mkver has quit [Ping timeout: 258 seconds]
<Daemon404> BBB, nicolas can never apologize, because it would be admitting he was wrong about somethin.g
<jamrial> we haven't even voted a new CC
<BBB> old CC is still in place
<elenril> we have no rules for sanctioning people
<BBB> but nicolas needs to soften his tone, at the very least
<Daemon404> cant wait for "CC is too old, i dont recognize ther authorit"
<BBB> :)
<jamrial> Daemon404: nicolas already did that
<j-b> it's been already done
<Daemon404> lmao
<Daemon404> life == satire
<Daemon404> got it
<j-b> that's exactly why I can't do any warnings anymore.
<BBB> ok I accept my nomination for the CC after all
<j-b> BBB: good.
<BBB> now get on and vote
<elenril> vote starts tomorrow
<j-b> the GA vote starts in 2 hours
<j-b> elenril: did you give a TZ?
<BBB> I wouldn't know, I'm not eligible to vote ;)
<j-b> or what is UTC?
<BBB> I'm asking to be added :-p
<elenril> not for starting to vote
<Daemon404> BBB, you can vote on the GA because youre on the old GA
<j-b> BBB: you are eligible to vote, for the old one.
<jamrial> BBB: commit cosmetics :p
<elenril> cosmetics explicitly don't count
<Daemon404> maybe we need a dependency manager for votes
<jamrial> curses
<elenril> quick, commit other people's patches
<elenril> those do count, according to a literal interpretation
<j-b> Authorship
<elenril> j-b: not what it says
<j-b> it should be changed, then.
<elenril> i agree
<elenril> nevertheless, what it says is "Contributors are considered "active contributors" if they have pushed more than 20 patches in the last 36 months in the main FFmpeg repository"
<courmisch> oooh, I should change to different personas, so I can vote multiple times
MisterMinister has joined #ffmpeg-devel
<j-b> courmisch: lol
<elenril> huh, it doesn't actually say anything about cosmetics
<j-b> yes, one could courmisches (yes, it's a word) its way into the GA
<Daemon404> just commit broken code and then slowly fix in in 10 commits
lemourin has quit [Quit: The Lounge - https://thelounge.chat]
<durandal_1707> Daemon404: that elenril doing all the time
<elenril> darn, foiled again
kurosu has quit [Quit: Connection closed for inactivity]
<Daemon404> no the tinfoil is worn by others
MrZeus__ has joined #ffmpeg-devel
<j-b> Paul does it all the time too
lemourin has joined #ffmpeg-devel
<durandal_1707> learned that from President
MrZeus has joined #ffmpeg-devel
<j-b> durandal_1707: don't badmouth Kieran, please.
MrZeus_ has quit [Ping timeout: 260 seconds]
<durandal_1707> j-b: where?
MrZeus__ has quit [Ping timeout: 255 seconds]
lemourin has quit [Client Quit]
<Daemon404> woosh
lemourin has joined #ffmpeg-devel
<BBB> hm... I was not on the list so I don't think I can vote
<BBB> there was a list of people who should get an email
<BBB> I wasn't on there
<Daemon404> idea: shut down ffmpeg for a month so everyone can touch grass.
<Daemon404> maybe two. or six.
<elenril> BBB: did you vote 3 years ago?
<durandal_1707> even 3yr ago BBB had 0 new commits in counter
lemourin has quit [Quit: The Lounge - https://thelounge.chat]
lemourin has joined #ffmpeg-devel
lemourin has quit [Quit: The Lounge - https://thelounge.chat]
lemourin has joined #ffmpeg-devel
mkver has joined #ffmpeg-devel
<BBB> grass is for nerds. we experts only touch pixels
goldsoft has quit [Quit: Connection closed]
goldsoft has joined #ffmpeg-devel
goldsoft has quit [Client Quit]
goldsoft has joined #ffmpeg-devel
<jkqxz> Grass is just that annoying-to-encode texture which some people like to draw on horizontal surfaces when rendering "sport" videos.
<jkqxz> It's pixels like everything else, don't worry.
<haasn> BtbN: I got one yeah
<haasn> The cheapest one they produce
<haasn> …
<haasn> And from amd I got their flagship model
<haasn> These two companies have some differences indeed
<cone-241> ffmpeg Kieran Kunhya master:7d497a1119b1: libavcodec/mpeg12: Remove "fast" mode
<cone-241> ffmpeg Kieran Kunhya master:2532e832d277: libavcodec/mpeg12: Reindent
markh has quit [Remote host closed the connection]
odrling has quit [Remote host closed the connection]
odrling has joined #ffmpeg-devel
odrling has quit [Remote host closed the connection]
derpydoo has joined #ffmpeg-devel
<BBB> jkqxz: best thing to do is blur the crap out of it. great for psnr metrics. winning!
<Lynne> not the libaom way
<Lynne> libaom way is to blur it until you get DCT basis functions sort of resembling grass
<Lynne> and then you make sure they're very very visible by blurring everything else to DCs
odrling has joined #ffmpeg-devel
markh has joined #ffmpeg-devel
odrling has quit [Remote host closed the connection]
odrling has joined #ffmpeg-devel
durandal_1707 has quit [Ping timeout: 245 seconds]
durandal_1707 has joined #ffmpeg-devel