Automated release: version and notes generated from pull requests merged since 2.21.1.
Changed
silence.py --margindefaults to 0.2 s, up from 0.15 s. Replies on X said strong cuts clipped breaths and weak word endings. The padding left on each side of every cut grows by 50 ms;--margin 0.15keeps the old cut. A template that sets its own margin keeps it.
Added
silence.py --timemap FILEwrites the source-to-output time map of a silence cut: every kept range with its source and output times, and every removed range with the output time it was cut at. It is JSON, or CSV for a.csvpath, and an existing map is refused without--overwrite. The map is written only after the cut it describes, and never onto the input, the cut or the--edlfile (timemap_path_taken,output_is_input). The silence-removal demo now writes one. It was requested on X for re-timing subtitles and finding a removed sentence again.silence.pyreports the level at every cut.cut_pointsgives the RMS 50 ms either side of each cut on the source, on the loudest channel (null, with acut_levels_unmeasuredwarning, when the source cannot be decoded for it).cut_warningsand acut_near_soundwarning name the cuts whose level is within 10 dB of--threshold, where a breath or a word's weak tail may have been cut. A probe can say the output's parameters are right, not where a cut landed; this points to where to listen.verifiedis unchanged.- README: what
verifiedcan and cannot tell you, and FAQ answers on how stable silence removal and captions are and on HDR-to-SDR conversion. The padding asked for around cuts already exists as--margin(now 0.2 s each side), and--speech-awarekeeps breaths inside a sentence; the FAQ and the scripts reference now say so. - Add silence.py --timemap and the level at every cut; default --margin 0.2 s; README says what verified cannot tell (#376)
What's Changed
- Add silence.py --timemap and the level at every cut; README says what verified cannot tell by @kajisho5 in #376
Full Changelog: v2.21.1...v2.22.0