sudacode/SubMiner
1
0
Fork 0
mirror of https://github.com/ksyasuda/SubMiner.git synced 2026-09-11 05:16:27 -07:00
Code Issues Packages Projects Releases Wiki Activity

fix(mining): ignore stale captions and improve waveform contrast

Browse Source 
- Use active parsed cue text and timings for overlay mining
- Normalize waveform levels against the clip's noise floor
This commit is contained in:
sudacode
2026-09-02 18:22:03 -07:00
parent f90ee78204
commit c85db5e07e
8 changed files with 145 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files
changes/media-timing-review.md
+1 -1
View File
@@ -1,5 +1,5 @@
type: added
area: mining
- Added optional pre-generation timing review for word, sentence, and audio cards with a compact speech-weighted waveform, clearly labeled mined-line boundaries, drag and keyboard adjustments, audio preview with a sweeping playhead, exact screenshot and AVIF timing, cancellation choices that include keeping a card without media, and a session-only runtime toggle.
- Added optional pre-generation timing review for word, sentence, and audio cards with a compact speech-weighted waveform that flattens steady background noise so dialogue edges are easy to see, clearly labeled mined-line boundaries, drag and keyboard adjustments, audio preview with a sweeping playhead, exact screenshot and AVIF timing, cancellation choices that include keeping a card without media, and a session-only runtime toggle.
- The timing review can pull any number of previous and next subtitle lines onto the card: `P`/`N` (or the Prev/Next steppers) add lines one at a time, Shift removes them, the sentence preview highlights exactly what the card will contain, and the clip range and line boundary markers on the waveform follow the added lines automatically.
changes/mining-stale-caption-row.md
+4
View File
@@ -0,0 +1,4 @@
type: fixed
area: mining
- Mining from the overlay no longer copies a lingering row of the previous broadcast caption onto the card sentence. The mined line and its clip timing now come from the parsed cue the overlay is showing, falling back to mpv's live text only when the cue list cannot explain it.