Update a table read's recording with the lines changed since it was made (free; TWO CALLS, quote first).

Brings the read's finished recording up to the read as it is now. The read knows what each line of its recording was spoken from — its voice, its words and its delivery (emotion, speed, volume, acting direction) — and the pause after it. A line whose words, voice or delivery changed since (a line edit, refine_performance) is spoken again by its voice's provider; every other line is kept byte for byte; a line whose only change is the pause after it is not spoken at all — the track is stitched again with the new pause; a line the read no longer speaks is stitched out. Then the read is mixed again with its music and sound, and the old recording keeps playing until the new one lands. FREE: 0 credits however many lines change — nothing is held, charged or booked, and credits is always 0. mode says what it would do: respeak, restitch (nothing spoken), remix (only the music or sound moved), record (no recording yet) or current (nothing to update). A recording with lines said over each other was mixed into one track, so no line can be cut from it (spliceable: false): every line is made again. TWO CALLS: sent without confirmed it QUOTES and changes nothing (quotedOnly: true, credits 0, mode, changedLines, respokenLines, pauseChanges, removedLines, keptLines, the changed lines, balanceAfter, sufficient, and a note to show the writer); sent again with confirmed: true it does it. Poll GET /artifacts/{artifactId} until manifest.audio.finalize.status is complete. Refused before the quote with 409 cast_voice_precondition_failed (details.reason): finalize_running (the recording is being made right now), no_recording (a read with no table-read artifact yet). 409 voice_retired when a line it must speak again is in a voice its provider retired. Requires Idempotency-Key. Scope: artifact:publish.