0

I'm not sure when this started happening to me, but I think it was with the advent of release 6.0, that when I use the "tpad" filter, like so:

-filter_complex "[0:0]tpad=start_duration=1:stop_duration=5:color=black[v0]" -map "[v0]"

the padded frames are green, regardless of what color I specify. I've tried this with mp4's, mkv's and avi's, and have gotten the same results. It's also been the same if I use the build from rpm-fusion-updates (6.0-11.fc38, libavfilter 9.3.100 / 9.8.101), or if I build it myself from the source (currently N-111066-gd26ea03f60, libavfilter 9.8.101 / 9.8.101).

It's happening on several different machines, too. Is anyone else seeing this? I'd like to confirm it's not just me before submitting a bug. I'm not seeing one open for this.

3
  • Doesn't happen here, with 6.0. Does happen, with git master.
    – Gyan
    Jun 18, 2023 at 3:58
  • The regression happened on May 22, likely commit 86b252ea9d
    – Gyan
    Jun 18, 2023 at 4:12
  • Have you tried other syntax like color=black or even omitting color for leaving the black default.
    – harrymc
    Jun 18, 2023 at 12:26

2 Answers 2

0

I have noticed the same issue under Windows 10. Switching to the below version of Ubuntu - no problem, no green screen, the default black was applied automatically. So, such workaround works.

Description: Ubuntu 20.04.6 LTS, Release: 20.04

Used Windows10 ffmpeg version: N-111283-g9b6d191a66-20230628

Used Ubuntu ffmpeg version: 4.2.7-0ubuntu0.1

Regards - Pawel.

2
  • As it’s currently written, your answer is unclear. Please edit to add additional details that will help others understand how this addresses the question asked. You can find more information on how to write good answers in the help center.
    – Community Bot
    Jun 29, 2023 at 10:58
  • The reason it does not happen with the Ubuntu version is that that is 4.2.7. The issue was introduced in 86b252ea9dee18006910e30646ad1067f2d1323f, currently not present in any release.
    – ePirat
    Nov 7, 2023 at 22:48
0

This is a regression from FFmpeg commit 86b252ea9dee18006910e30646ad1067f2d1323f that I made a while ago. The bug report for this regression is at https://trac.ffmpeg.org/ticket/10621.

I've submitted a fix for this to the mailing list and it will probably get merged soon.

You can work around the issue by specifying the start and/or stop option (depending which you need) with a dummy value like 1, while still specifying the start_duration and/or stop_duration, which will take precedence.

So given your example above, the following should work around the issue:

-filter_complex "[0:0]tpad=start=1:start_duration=1:stop_duration=5:color=black[v0]" -map "[v0]"

(When both start_duration and stop_duration are used, just giving either start or stop with a dummy value is enough to make the broken check happy.)

Sorry for this regression!

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .