get transcript
Every spoken word on the timeline with its position in the cut, as frame numbers at the timeline’s rate, the positions remove_words takes. granularity words: [index, text, from_frame, to_frame], 10000 a page; sentences: [first_word_index, text, from_frame, to_frame], 1000 a page, first_word_index pointing into words mode. A full page carries next_start_frame. Indices are positions in the whole transcript and stable across pages. speakers adds who was talking on each row, from what the speakers tool found. An empty result usually means untranscribed, not silent: transcribe_media with timeline_id gives the clips their words. For one library item, transcribe_media with ref. Free.

