BtbN changed the topic of #ffmpeg to: Welcome to the FFmpeg USER support channel | Development channel: #ffmpeg-devel | Bug reports: https://ffmpeg.org/bugreports.html | Wiki: https://trac.ffmpeg.org/ | This channel is publically logged | FFmpeg 7.0 is released
minimal has quit [Quit: Leaving]
Dagger has quit [Ping timeout: 276 seconds]
glars has joined #ffmpeg
Muimi__ has quit [Quit: Going offline, see ya! (www.adiirc.com)]
Dagger has joined #ffmpeg
georgereynolds80 has joined #ffmpeg
georgereynolds8 has quit [Ping timeout: 252 seconds]
georgereynolds80 is now known as georgereynolds8
Dagger has quit [Ping timeout: 244 seconds]
\\Mr_C\\ has joined #ffmpeg
Dagger has joined #ffmpeg
SystemError has quit [Ping timeout: 260 seconds]
SystemError has joined #ffmpeg
Tinos has quit [Remote host closed the connection]
ppw has joined #ffmpeg
JanC_ has joined #ffmpeg
JanC is now known as Guest9089
Guest9089 has quit [Killed (molybdenum.libera.chat (Nickname regained by services))]
JanC_ is now known as JanC
<ppw>
suggestions for what to do here? [in#1/rawvideo @ 0x5ae955fc5cc0] Thread message queue blocking; consider raising the thread_queue_size option (current value: 8)
SystemError has quit [Remote host closed the connection]
Tinos has quit [Remote host closed the connection]
Muimi__ has joined #ffmpeg
down200 has quit [Ping timeout: 245 seconds]
down200 has joined #ffmpeg
rvalue has quit [Read error: Connection reset by peer]
rvalue has joined #ffmpeg
lucasta has quit [Remote host closed the connection]
memset_ has quit [Remote host closed the connection]
memset has joined #ffmpeg
anticw has quit [Remote host closed the connection]
anticw has joined #ffmpeg
lavaball has joined #ffmpeg
<ta71>
this may not be ffmpeg specific but is there a way to create a sample video like just a single character A in big for whole frame for 5 seconds, B for 5 seconds etc
AbleBacon has quit [Read error: Connection reset by peer]
rv1sr has joined #ffmpeg
vlm has joined #ffmpeg
squeaktoy has quit [Ping timeout: 260 seconds]
Blacker47 has joined #ffmpeg
xx has joined #ffmpeg
HerbY_NL has joined #ffmpeg
Tinos has joined #ffmpeg
echelon has quit [Quit: echelon]
jagannatharjun has joined #ffmpeg
zzarr has joined #ffmpeg
martylake has quit [Ping timeout: 248 seconds]
martylake has joined #ffmpeg
memset has quit [Remote host closed the connection]
memset has joined #ffmpeg
glars has quit [Remote host closed the connection]
glars has joined #ffmpeg
HerbY_NL_ has joined #ffmpeg
HerbY_NL_ has quit [Client Quit]
HerbY_NL has quit [Ping timeout: 260 seconds]
beaver has quit [Remote host closed the connection]
yawkat has quit [Ping timeout: 244 seconds]
beaver has joined #ffmpeg
yawkat has joined #ffmpeg
glars has left #ffmpeg [ERC 5.6.1-git (IRC client for GNU Emacs 31.0.50)]
zzarr has quit [Quit: Leaving]
yawkat has quit [Ping timeout: 252 seconds]
vincejv has quit [Remote host closed the connection]
Tinos has quit [Remote host closed the connection]
beaver has quit [Remote host closed the connection]
beaver has joined #ffmpeg
Kei_N has joined #ffmpeg
<DeHackEd>
ta71: yes, you can build a video using only video filters...a blank image to start, a text drawing on top of that, etc... there are also some testsrc filters to generate moving test patterns
fling_ has joined #ffmpeg
fling has quit [Remote host closed the connection]
peterLustig67 has joined #ffmpeg
fling_ is now known as fling
yawkat has joined #ffmpeg
Blacker47 has quit [Quit: Life is short. Get a V.90 modem fast!]
iliv has quit [Ping timeout: 252 seconds]
minimal has joined #ffmpeg
yans has joined #ffmpeg
wastekoko has quit [Read error: Connection reset by peer]
wastekoko has joined #ffmpeg
Blacker47 has joined #ffmpeg
kasper93_ has joined #ffmpeg
kasper93 has quit [Ping timeout: 260 seconds]
coldfeet has joined #ffmpeg
alexherbo2 has joined #ffmpeg
evilscreww has quit [Quit: Leaving]
peterLustig67 has quit [Ping timeout: 256 seconds]
zzarr has quit [Quit: Leaving]
travisghansen has joined #ffmpeg
Muimi__ has joined #ffmpeg
coldfeet has quit [Remote host closed the connection]
AbleBacon has joined #ffmpeg
vincejv has joined #ffmpeg
bigtimeconfused has joined #ffmpeg
<bigtimeconfused>
hi there. has anyone out there had success using vulkan / cuda / nvdec / nvenc interop using any of the ffmpeg 7 releases? works for me without -hwaccel cuda -hwaccel_output_format cuda, not so much with.
<bigtimeconfused>
i've tried linux, windows, different versions of the CUDA SDK, NVIDIA drivers, Vulkan SDK, etc ... unless something fundamental has changed in the way i should be structuring commands in 7, something seems to be up. what was working fine and dandy in 5.1.3 is now toast.
Vonter has quit [Quit: WeeChat 4.3.5]
<BtbN>
The Vulkan code in old ffmpeg was super busted and has been completely rewritten and is still in the process of being worked on/completed.
<BtbN>
Right now the Vulkan stuff is always half-broken, cause of wonky drivers, vendor-differences, under-documentation, ...
<BtbN>
best wait 5 years or so for actually using it
<BtbN>
(Also, the CUDA SDK is 100% irrelevant. FFmpeg does not use it in any way)
<bigtimeconfused>
that's what i've been hearing :( There was a patch that fixed things in 5.1 - 6 i believe, but then a change was made to hwcontext_vulkan that made it no longer usable in 6.1+
<bigtimeconfused>
I actually use 5.1.3 in a large production environment with that patch and it works great
<BtbN>
It was rewritten from scratch, cause it was unuseably broken and Vulkan itself basically completely changed
<bigtimeconfused>
Just getting kind of old and looking to upgrade.
<BtbN>
Vulkan in 5.x was near useless. Never heard of it working at all
<JEEB>
since I know some things work, you will have to actually detail what use case doesn't work any more with a pastebin or so
<JEEB>
and link it here
<JEEB>
otherwise we're just discussing really generic stuff
<bigtimeconfused>
sure. just a sec
<JEEB>
and possibly completely unrelated to whatever you're seeing
<bigtimeconfused>
these are just examples. it doesn't matter what the input is - all NVDEC decodable and works fine w/ 5.1.3 with the patched hwcontex_vulcan - same result. seems like something is trying to force that auto scale into the filter chain, and then the -40 (Function not implemented) and -22 (Invalid argument)
<bigtimeconfused>
results are slightly different w/ linux but basically the same.
<JEEB>
1) `-v verbose` is the last log level that is not spammy, and it does log various filter chain things etc, so I recommend utilizing it
<bigtimeconfused>
*hwcontext_vulcan
<JEEB>
2) `-noauto_conversion_filters` disables auto-insertion of the scale filter
<JEEB>
and yea, when hitting an issue I recommend first testing current master since any bug fixes or improvements get handled there first, and master is supposed to be all green @ FATE (https://fate.ffmpeg.org/) also in various cases fixes either aren't or can't be easily pulled into some release branch(es).
<bigtimeconfused>
Gotcha - yeah, still no joy with -noauto_conversion_filters. Fails, but a bit differently.
<JEEB>
yes, it should actually fail with the actual error as opposed to when it attempts to auto-insert. coupled with `-v verbose` it should also log the configured state of at least some filters
<BtbN>
something unrelated to keep in mind: There is no actual way to map/derive frames between CUDA and Vulkan
<BtbN>
so every hwmap is basically a copy
<JEEB>
right, so the hwupload is not there yet
<JEEB>
oh, or hwupload is but it's not a direct import of surfacE?
<JEEB>
but rather a VRAM memcpy
<BtbN>
Yeah, you can't just map, it's a memcpy
<JEEB>
gotcha
<JEEB>
vulkan did get extensions for d3d11 import etc, not sure if they have that for cuda yet
<BtbN>
Nvidia is extremely efficient with those though
<JEEB>
yea
<JEEB>
bigtimeconfused: anyways feel free to post the log without auto conversions and with verbose
<bigtimeconfused>
will do in just a sec, just wanted to find the patch that magically made all of this stuff work in 5.1.3
<bigtimeconfused>
thanks. i've been banging my head on this for weeks, lol. this is my last stop before filling out a bug report, basically
<JEEB>
> trim_in_0:0
<JEEB>
auto-insertion of trim by fftools?
figboot has joined #ffmpeg
Ram-Z_ has joined #ffmpeg
<JEEB>
did not think trim cared about pix_fmts...
sm1999_ has joined #ffmpeg
rvalue has quit [Ping timeout: 252 seconds]
rvalue- has joined #ffmpeg
<JEEB>
bigtimeconfused: while I'm checking the logic can you check you get the same with BtbN's master build that I linked earlier?
<bigtimeconfused>
happy too, sure. appreciate it
sm1999 has quit [Ping timeout: 246 seconds]
bigfoot547 has quit [Ping timeout: 246 seconds]
Ram-Z has quit [Ping timeout: 246 seconds]
<JEEB>
ok, so I have just completely missed the (a)trim stuff
<JEEB>
that is as old as 2013
rvalue- is now known as rvalue
<JEEB>
and since it isn't doing anything to those frames it failing between the input and it sounds really wonky
<JEEB>
ok no, so first failure is on the audio side since auto-conversions are disabled. you basically need a `-af aresample` to just have something that converts from input to output config (as they differ)
<JEEB>
as AAC decodes as planar float 44.1kHz and the default audio encoder for null muxer is pcm_s16le
<JEEB>
so a conversion between those two needs to be allowed
<JEEB>
(by adding an aresample filter)
<JEEB>
(or just doing `-map 0:v` which will just map all video streams from input 0 to output
<JEEB>
and thus skipping audio
<bigtimeconfused>
results w/ BtbN's version look, to my untrained eyes at least, basically the same
<JEEB>
yup, so remove the audio from the equation just in case
<JEEB>
by just mapping the video
<bigtimeconfused>
would -an accomplish the same thing?
<JEEB>
yes and no. it would make the auto-selection not select audio, but it would still pick other types of stuff. I like being explicit. that disables auto-selection and picks only what gets mapped
<bigtimeconfused>
never tried it, as the change to hwcontext_vulkan in 6.1 specifically made that patch i mentioned which fixed things stop applying.
<JEEB>
well that patch seemed linux specific (mentions drm modifiers which are a linux thing), and not generic avfilter or ffmpeg cli related stuff
iliv has joined #ffmpeg
<JEEB>
the BtbN link has current state of 6.1 branch build there as well
<JEEB>
so should be simple to test
jemius has joined #ffmpeg
<JEEB>
and the error we're getting doesn't seem to be pointing at all towards the hwcontext
<JEEB>
or well, at this point you are getting
<bigtimeconfused>
Yes, that patch was applied and used in linux.
<bigtimeconfused>
will try that 6.1
<JEEB>
cheers
stolen has joined #ffmpeg
<JEEB>
since the trim stuff has been there since 2013, it suddenly failing means that something somewhere changed and it would be nice if that range could be limited somewhat
<JEEB>
and 5.1-latest? if I go by what you are saying that in theory should work
<JEEB>
ah no, libplacebo is disabled on 5.1
bitblit has joined #ffmpeg
<JEEB>
but hwupload to vulkan from cuda should be a thing
<JEEB>
since we're not even getting that far as libplacebo
<bigtimeconfused>
and now you see why i've been going nuts over this for 3 weeks ... lol. as i recall, it was that patch + one magic version of libplacebo
<bigtimeconfused>
however, think libplacebo could be taken out of the mix. i bet you any vulkan filter would fail.
<bigtimeconfused>
*i think
<JEEB>
yea
<JEEB>
esp. since the error is way before that
<viric>
damn vaapi...
<viric>
I don't know why it picks a wrong default frame format
<viric>
I'm trying to make sense of the hw sw_formats and hw_formats
microchip_ has quit [Remote host closed the connection]
emanuele6 has quit [Ping timeout: 276 seconds]
microchip_ has joined #ffmpeg
emanuele6 has joined #ffmpeg
SakuraChan has quit [Remote host closed the connection]
kasper93_ is now known as kasper93
bigtimeconfused has quit [Quit: Client closed]
bigtimeconfused has joined #ffmpeg
microchip_ has quit [Quit: There is no spoon!]
microchip_ has joined #ffmpeg
Sakura`Kinomoto has joined #ffmpeg
bigtimeconfused has quit [Quit: Client closed]
kmikita has quit [Ping timeout: 252 seconds]
militantorc has quit [Ping timeout: 244 seconds]
waleee has joined #ffmpeg
moviuro has quit [Ping timeout: 265 seconds]
moviuro has joined #ffmpeg
bigtimeconfused has joined #ffmpeg
Kei_N has quit [Ping timeout: 276 seconds]
Kei_N has joined #ffmpeg
memset has quit [Remote host closed the connection]
memset has joined #ffmpeg
Narrat has joined #ffmpeg
kmikita has joined #ffmpeg
chiselfuse has quit [Remote host closed the connection]
chiselfuse has joined #ffmpeg
Narrat has quit [Quit: They say a little knowledge is a dangerous thing, but it's not one half so bad as a lot of ignorance.]
Livio has joined #ffmpeg
Tinos has joined #ffmpeg
Rena has quit [Quit: $WITTY_QUIT_MESSAGE]
Blacker47 has quit [Quit: Life is short. Get a V.90 modem fast!]