[FFmpeg-devel] [PATCH 0/1] Fix tmix on discontinuities

[FFmpeg-devel] [PATCH 0/1] Fix tmix on discontinuities

[Ignacy Gawędzki]

A simple example of the bug:

  ffplay -f lavfi color=c=gray -vf 'tmix=enable=gt(t\,1)'

After one second, the color is wrong (green in my case), not the
expected gray.

It seems it is due to the fact that parse_weights determines that
the weights are all 1 and FAST_TMIX_SLICE should be used instead of
MIX_SLICE.  But if the filter is not running from the very start with
these weights, FAST_TMIX_SLICE will not work properly.

Besides, should FAST_TMIX_SLICE not be used with float types either,
as addition on floats is not associative and rounding errors will
accumulate?

Ignacy Gawędzki (1):
  avfilter/f_mix: disable fast on discontinuities

 libavfilter/vf_mix.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

-- 
2.48.1
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Re: [FFmpeg-devel] [PATCH 0/1] Fix tmix on discontinuities

[Michael Niedermayer]

[-- Attachment #1.1: Type: text/plain, Size: 1282 bytes --]

Hi Ignacy

On Sun, Jul 06, 2025 at 08:30:09PM +0200, Ignacy Gawędzki wrote:
> A simple example of the bug:
> 
>   ffplay -f lavfi color=c=gray -vf 'tmix=enable=gt(t\,1)'
> 
> After one second, the color is wrong (green in my case), not the
> expected gray.
> 
> It seems it is due to the fact that parse_weights determines that
> the weights are all 1 and FAST_TMIX_SLICE should be used instead of
> MIX_SLICE.  But if the filter is not running from the very start with
> these weights, FAST_TMIX_SLICE will not work properly.

It should be possible to fix fast mode so it resets its state when
teh filter re-enables

But as long as thats not done, i think its ok to fix it by disabling fast
mode


> 
> Besides, should FAST_TMIX_SLICE not be used with float types either,
> as addition on floats is not associative and rounding errors will
> accumulate?

yes, the filter is wrong to use this with floats
though if there was a state "reset" for re-enabling the filter
that could also be used for float periodically

thx

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Does the universe only have a finite lifespan? No, its going to go on
forever, its just that you wont like living in it. -- Hiranya Peiri

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 251 bytes --]

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".