Stage 1: manual trim from raw audio → Stage 2: preprocess manual segments → AI transcription → review → export.