Repository navigation
Export: split at timeline markers into named chapters from one render (#527) - #764
Conversation
…ender (#527) Forced keyframes at the markers and the segment muxer cut one render, so feedback layers stay continuous across chapters. Segment split times sit half a frame early; at the exact time x264 skipped a cut. AMF ignores forced keyframes and is not picked for chapters. The bench driver now holds panel config updates, which overwrote the first job's settings. Co-authored-by: caganturgut <271973834+caganturgut@users.noreply.github.com>
Co-authored-by: caganturgut <271973834+caganturgut@users.noreply.github.com>
|
CaYaDraft — pull request Summary Files README.md (modified, +5/-3) README.tr.md (modified, +6/-3) ROADMAP.md (modified, +20/-2) src/admin/admin.js (modified, +7/-0)
+… src/main/main.js (modified, +64/-1)
Note Stopping here. I will not merge, push, or change this pull request until you ask.
|
Closes #527.
What changed
01 - Intro.mp4,02 - Drop.mp4..partialand renamed only when the export succeeds, as with frame sequences. If the file count does not match the plan, the job fails rather than silently missing a chapter.chaptersboolean.Two traps, both measured
-forced-idr 1, Quick Sync, or the CPU.Measured
In the app, 5 s of audio, markers at 2.0 s ("Nakarat") and 3.5 s ("Son: Bitiş"), 30 fps:
01 - Başlangıç,02 - Nakaratand03 - Son Bitiş.Also fixed: the bench driver's first job
The panel sends its config while it loads, and that overwrote the first bench job's settings. That is where the odd first-job hash noted in #760 came from, and it is why chapters first saw no markers. The driver now holds panel updates during the bench, the same way the self-test does.
Verification
npm test: 3145/3145.tests/export-chapters.test.jscovers the plan, names, arguments, the AMF exclusion, main-process wiring, the bench hold, and the panel and MCP. The existing smoke-hold and export-job tests were updated for the second hold caller and the extra error path.