Re-voice one line of a table read at a new speed or with a new acting direction (free; TWO CALLS, quote first).

Speaks line entryIndex again from its voice's provider with a new speed, a new actingDirection, or both (the line is spoken once). The speed is sent as that provider's OWN speed parameter (Cartesia generation_config.speed 0.6–1.5, ElevenLabs voice_settings.speed 0.7–1.2, Deepgram speed 0.7–1.5) — never a time-stretch — and brought inside the provider's range; the answer says the range (range) and whether it did (clamped). The direction is free text in the writer's words, sent through the provider's OWN expressive control: OpenAI instructions, ElevenLabs Eleven v3 and later as a leading audio tag, and Cartesia sonic-3 and newer as one emotion from the voice's list and a pace nudge, chosen from the direction's words by the mapping step when the line is spoken — GET /scripts/{scriptUploadId}/lines/{entryIndex} then shows it as sentAs (direction says which channel); an empty string or null takes it off. Every other line keeps its recording byte for byte, then the read is mixed again with its music and sound. FREE: a line's speed and direction cost no credits — nothing is held, charged or booked, and credits is always 0. Three outcomes (mode), all quoted first: revoice — the read has a finished recording, and the line is spoken again; restore_take — the line had a take with this speed and direction before (the take it has is kept whenever it is re-voiced), and it is brought back; save — the read has no recording yet: they are saved and the line is spoken that way when the read is recorded. TWO CALLS: sent without confirmed it QUOTES and changes nothing (quotedOnly: true, credits 0, mode, changes, speed, previousSpeed, range, actingDirection, previousActingDirection, keptLines, 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): provider_has_no_speed (an OpenAI or Hume voice, or a model or voice with no speed setting — Eleven v3, Cartesia before sonic-3, a Deepgram Aura-1 voice or a Cartesia professional clone — named in details.model and details.noSpeed: recast the character first), provider_has_no_direction (a Deepgram, Hume, or OpenAI tts-1 voice, ElevenLabs before Eleven v3, or a Cartesia model before sonic-3 — none reads a written direction, and Cartesia before sonic-3 has no delivery controls at all — details.model, details.noDirection), same_speed, same_direction, nothing_spoken, no_voice, modified_take (the line plays a voice effect or a performance take), finalize_running, not_spliceable / not_in_recording (a recording that cannot be cut line by line), line_changed (the line was edited since it was priced). 409 voice_retired when the line's voice is one its provider retired. Requires Idempotency-Key. Scope: artifact:publish.