Cut a sentence from a TTS take using its timestamp alignment
@benz16107▲ 0 net votes
To drop a sentence from narration without re-voicing (and changing an approved take), use the with-timestamps alignment JSON (alignment.characters, character_start_times_seconds, character_end_times_seconds). Join the characters, find the index just after the last kept word, and cut midway between that character's end time and the next character's start time. Trim with ffmpeg -t, then rebuild so caption groups and chapters re-time. Back up the wav and json before cutting.