michaelni changed the topic of #ffmpeg-devel to: Welcome to the FFmpeg development channel | Questions about using FFmpeg or developing with libav* libs should be asked in #ffmpeg | This channel is publicly logged | FFmpeg 7.1 has been released! | Please read ffmpeg.org/developer.html#Code-of-conduct
<fflogger>
[newticket] Wallboy: Ticket #11466 ([avformat] FIFO Muxer: restart_with_keyframe does not work when drop_pkts_on_overflow false) created https://trac.ffmpeg.org/ticket/11466
System_Error has quit [Remote host closed the connection]
<fflogger>
[newticket] Wallboy: Ticket #11467 ([avformat] FIFO Muxer: restart_with_keyframe recovers on audio packets) created https://trac.ffmpeg.org/ticket/11467
System_Error has joined #ffmpeg-devel
ngaullier has joined #ffmpeg-devel
System_Error has quit [Remote host closed the connection]
System_Error has joined #ffmpeg-devel
<michaelni>
haasn, cosminaught, if you want to add a project to gsoc 2025. You should do it soon because the number of slots we receive may depend on how many ideas we have on the page when google reviews it.
<michaelni>
ePirat, are you interrested in mentoring a gsoc project ?
<michaelni>
the slot number could also depend on how many unique mentors we have, i have no idea, but more is certainly better
<pross>
i'd like for there to be a 'silk v3' decoder in ffmpeg. there is reference decoder source code. but i don't have the skills to mentor such an unfathomably complex project.
<michaelni>
maybe theres a simpler project you want to mentor ?
<michaelni>
maybe an encoder for some codec that already has a simple decoder?
<pross>
done.
<Lynne>
silk v3?
<pross>
vp6 encoder
<pross>
done as-in added to the wiki :)
<Lynne>
oh, doesn't look like k.vos was involved in silk v3, boring
<pross>
silk v3 is used by weixin/wechat for voice message compression
<Lynne>
weird that they wouldn't just use opus
cone-858 has joined #ffmpeg-devel
<cone-858>
ffmpeg Martin Storsjö master:b053f1173d97: random_seed: Reorder if clauses for gathering entropy
<cone-858>
ffmpeg Martin Storsjö master:19a2d2617711: random_seed: Improve behaviour with small timer increments with high precision timers
Guest59 has joined #ffmpeg-devel
xvaclav has quit [Quit: Ping timeout (120 seconds)]
xvaclav has joined #ffmpeg-devel
Guest59 has quit [Quit: Client closed]
SecretLeader has joined #ffmpeg-devel
<kierank>
SecretLeader: are you enjoying git thread
<kierank>
it's really something else
<SecretLeader>
there so many mails, that i read every 10th
<Lynne>
from the genius that brought you subtitles in avframe done wrong
<kierank>
federation solves all problems
<kierank>
>narrator: it solves none
<Lynne>
if you've never used anything with federation, why do you care?
<SecretLeader>
all lazy devs here should be banned
<kierank>
because shouting about it as the solution to all problems when it solves nothing convinces nobody
<Lynne>
you don't know that
<Lynne>
how can you if you've never even bothered to research
<kierank>
it literally solves your pet problem
<kierank>
and nobody elses
<Lynne>
good.
<SecretLeader>
"do you own research"
<kierank>
SecretLeader: yeah we should use some jank system where we can't fork, some weird git commands
<Lynne>
all you've been doing for the last N months is trying to find faults with forgejo
<kierank>
all you've done for the last year is say "federation"
<kierank>
providing literally 0 benefits
<Lynne>
but all you've ended up is parroting the same misconceptions you've held since you heard the name
<SecretLeader>
stop fighting over irrelevant things
<kierank>
forgejo is doing a great job of making itself look amazing right now
<Lynne>
as well as it should be
<kierank>
sarcasm is travelling well on the internet I see
<Lynne>
sarcasm is wasted on those who ignore it
<kierank>
I still don't understand how this Agit thing doesn't create infinite branches in the main repo
<kierank>
if they are not forks
<kierank>
not that this Agit thing is usable anyway
<Lynne>
agit should't even exist imo
<kierank>
so it's a moot point
<Lynne>
yeah
<Lynne>
its some deprecated system carried over from gittea that forgejo admits has a limited implementation
<kierank>
yet that's what is being proposed we use
<Lynne>
exactly, it shouldn't even be proposed
<SecretLeader>
forgejo is here to stay, use librempeg instead
<Lynne>
at the end of the day, regardless of where we end up or what we do, you'll need to ask an admin or wait until you can make a repository
<kierank>
not on github
<kierank>
:)
<SecretLeader>
on gitlab?
<kierank>
forjego, it's federated
<kierank>
brawndo, it's got electrolytes
<SecretLeader>
name is lame, but that is not important
<Lynne>
if we're moving to github, maybe we should move chat to discord too
<Lynne>
I'm only half-joking
<SecretLeader>
move everything to some random 3rd party service provider
<SecretLeader>
true federation
<kierank>
SecretLeader: isn't that the current case :)
<kierank>
everything is on a random third party provider
Lynne has quit [Read error: Connection reset by peer]
Guest59 has joined #ffmpeg-devel
<thardin>
could move to fossil. that way code, tickets and wiki are all handled via the SCM system
<thardin>
it's a really neat idea
<thardin>
how do you pronounce forgejo anyway?
<SecretLeader>
f o r g e t o
<wbs>
thardin: "forg(e)" + "eyo"
<thardin>
https://code.ffmpeg.org/ ought to list the repos immediately imo instead of negative space webslop
<thardin>
seems browsable without js at least. a plus for blind/semi-blind devs
<thardin>
forge-joe
Guest59 has quit [Quit: Client closed]
<SecretLeader>
ban javascript
<kierank>
ban librempeg
<kierank>
Powered by Forgejo Version: 10.0.1 Page: 3381ms Template: 205ms
<kierank>
>gitlab is slow
<kierank>
it's comparable in speed
<SecretLeader>
what?
<kierank>
gitlab and forjego
<SecretLeader>
github handless myriad users, and fogejo only hadles single one
<thardin>
3 seconds to generate a page sounds awful
<thardin>
is it running on some anemic VM?
<cone-858>
ffmpeg Anton Khirnov release/7.1:8cabfd922a72: lavc/hevcdec: unbreak WPP/progress2 code
<nevcairiel>
it is a VM, how fast it is no idea, what kind of page did you query?
<thardin>
Drivs av Forgejo Version: 10.0.1 Sida: 11446ms Mall: 2094ms
<thardin>
13½ seconds..
<nevcairiel>
hm yeah it seems pretty slow just on the main repository page, either its being bothered by web crawlers again or maybe some caching/indexing is running wrong
<nevcairiel>
my own Gitea (not quite forgejo, but similar enough) is pretty instant on a ffmpeg repo, but I seemingly managed to get the web crawlers to leave it alone =P
<SecretLeader>
compn: do you use mplayer audio filters?
<Compn>
yes, volnorm, and volume mostly
<Compn>
sometimes karaoke
<Compn>
probably some more am not thinking of right now
<kierank>
thardin: I prefer the loading ajax
<kierank>
compared to forejgo where you wait for the whole thing to load
<SecretLeader>
compn: there are very low number of users, can you use/try it, and tell your opinion?
<Compn>
SecretLeader, i watch a lot of films. unfortunately the audio is mixed improperly for playback in home theaters / home tvs. so the dialog is low, the explosions are loud. i think if someone made a hardware hdmi pass-through with volume normalization it would sell well.
<JEEB>
compn: that's just standardized downmixing at work.
<SecretLeader>
you cant get hdmi pass-audio and volume normalization at same time, unless you refilter and rencode full audio stream
<Compn>
yes, thats what i'm suggesting SecretLeader
<Compn>
real time filter for normalization, normal to standard...
<Compn>
not to itself
<JEEB>
but yes, thankfully loudnorm dynaudnorm etc do the job and at least I can insert those after downmix
<Compn>
i guess with more smart tvs, an hdmi passthrough wont help. the volume normalization has to go into the tv itself :\
<SecretLeader>
compn: with lavfi-preview you can open movies add filters processing you like and than watch it
<Compn>
SecretLeader, yes i know what lavfi-preview does
<Compn>
i'm just not reencoding much anymore, i dont have need for ffmpeg filters on a daily basis
<Compn>
i like your project though
<SecretLeader>
you can also change filter parameters at runtime, for filters that support it
<Compn>
stuff i'd use: a lavfi volume booster browser extension
tufei__ has quit [Remote host closed the connection]
<SecretLeader>
doesnt that need compilation?
tufei__ has joined #ffmpeg-devel
<Compn>
sure if you dont write it in javascript
<SecretLeader>
all this situation killed my motivation to finish RE of sonarc codec
Marth64 has joined #ffmpeg-devel
<arch1t3cht>
<@BtbN> With full approval only registration you completely kill off contribution though :/ - can confirm, my videolan account has been in approval limbo for a year
<BtbN>
I like the model Lynne(?) proposed on the ML, that there is a seperate repo, where you open an "Hey, I'm not a bot!" issue on.
<SecretLeader>
just use github
<BtbN>
That way you don't need to go via some separate channel to get approved
<Lynne>
one could probably write a bot which listens to emails and automatically approves the user
<Lynne>
at the end of the day it's all done via http requests
ngaullier has quit [Remote host closed the connection]
<SecretLeader>
https?
rvalue- has joined #ffmpeg-devel
rvalue has quit [Ping timeout: 265 seconds]
mkver has quit [Ping timeout: 252 seconds]
rvalue- is now known as rvalue
SecretLeader has quit [Quit: Client closed]
SecretLeader has joined #ffmpeg-devel
Tanay has quit [Ping timeout: 276 seconds]
<ramiro>
I want to add Spout support to FFmpeg <https://spout.zeal.co/>. Basically it's a way to share OpenGL/DirectX textures with other programs in real time.
<ramiro>
I'd start with writing output support, but I'm not quite sure where/how to write it.
<ramiro>
I know Anton dislikes the idea of adding stuff like that to libavdevice, so I guess it would be a lavfi video sink.
<ramiro>
But Spout has no synchronization mechanism, it's basically "call SendImage() and it will be shared right away with whoever calls ReceiveImage()".
<ramiro>
I guess adding synchronization mechanism to a video sink would also be a no-no, so I'm wondering if this shouldn't actually be an ffplay renderer instead.
<ramiro>
Any suggestions?
<Lynne>
on linux, pretty much everything can be exported as a dmabuf, which pipewire may route, and it has synchronization
<Lynne>
is there some niche that spout has or support from other clients?
<ramiro>
Lynne: on linux it's done and pretty straightforward, but on windows it's not as simple :/
<Lynne>
we support d3d11, which is pretty much the standard these days, isn't it?
<ramiro>
VJs use Spout to share visuals to Resolume and TouchDesigner
<Lynne>
all capture APIs afaik output d3d11
<Lynne>
ah
<ramiro>
Spout's SendImage() is a bit annoying in that it takes width and height, but it doesn't support stride like in FFmpeg. I might have to add that to Spout first to prevent having to pack the data first before sending it.
<ramiro>
apparently it has a region in shared memory where you can register the name of your sender and its directx/opengl texture id (or something like that). I'm wondering if it wouldn't be better to just implement this natively instead of using their dlls.
<ramiro>
Lynne: another question I wanted to ask you at fosdem but I didn't get the chance, would it be possible to get an opengl buffer and use that as the output to the decoder, instead of outputting to memory and then copying to the opengl buffer? or is this not possible/slow/bad design?
<Lynne>
nope
Guest59 has joined #ffmpeg-devel
<Lynne>
I don't think opengl code belongs in ffmpeg since we're a library, and opengl is notoriously global
<ramiro>
Lynne: thanks, that's the kind of quick reply I was hoping for :)