CalvinBackup/remove-ai-watermarks
1
0
Fork 0
mirror of https://github.com/wiltodelta/remove-ai-watermarks.git synced 2026-08-09 23:50:40 +02:00
Code Issues Packages Projects Releases Wiki Activity

Bound ffmpeg video encoder threads

Browse Source
This commit is contained in:
Victor Kuznetsov
2026-07-31 14:33:50 -07:00
parent 7fb0e65455
commit 166dab9d03
5 changed files with 22 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/module-internals.md
+6 -5
View File
@@ -149,11 +149,12 @@ regeneration. It centralizes container codecs, optional audio stream copying,
metadata/chapter policy, encode-failure reporting, and atomic same-directory
publication. Each mapped stream is allowed to reach its own end, so a copied
audio tail is not shortened to the frame-input duration.
Both the raw-BGR and timestamped-NUT stdin inputs disable ffmpeg probing before
`pipe:0`: their format is already explicit, and ffmpeg 6 on Linux can otherwise
wait for its normal analysis window while the producer blocks on a full pipe
before the source-audio input opens. Command-order regressions cover both stdin
modes; the real Linux full-clip CI job guards process completion.
Both the raw-BGR and timestamped-NUT stdin modes cap the video encoder at two
threads. A Linux full-clip trace showed ffmpeg creating dozens of worker stacks
and severely delaying frame-pipe ingestion on a constrained hosted runner. The
bounded codec pool avoids that scheduling collapse while leaving audio stream
copy independent. Command regressions cover both supported video codecs; the
real Linux full-clip CI job guards process completion.
`probe_video_encode_profile` reads the first source video stream with ffprobe
and preserves the supported properties that survive the 8-bit BGR boundary:
`yuv420p`/`yuv422p`/`yuv444p` chroma sampling, recognized color tags, encoder