Refine the delivery of all of one character's lines on a table read artifact.

Adjusts the performance — speed, volume, and cadence — across ALL of one character's lines on the live read backing a table_read artifact. How lines feel is their written acting directions (there are no emotion tags; on a Cartesia voice a direction is sent as one emotion from the voice's list): change one with POST /artifacts/{artifactId}/revoice-line. emotions and overrideExisting are refused with 400 validation_failed. Set character to a name, "NARRATOR" (all narration lines), or "ALL" (every voiced line). Speed and volume accept either an absolute value (speed, 1 the natural pace; volume 0.5–2.0) or a relative per-line multiplier (speedFactor/volumeFactor, e.g. 1.2 for +20%). Each line's speed is brought inside the range of the provider that speaks it (Cartesia 0.6–1.5, ElevenLabs 0.7–1.2, Deepgram 0.7–1.5), and a line whose voice takes no speed setting (OpenAI, Hume, Eleven v3, Cartesia before sonic-3, Deepgram Aura-1, a Cartesia professional clone) keeps its pace — speedUnsupportedLines counts them and speedNote says why. Volume clamps to 0.5–2.0. Provide at most one of each pair. There is no separate cadence field — express cadence as speed plus pauseAfterMs (breathing room after each line). No new revision — the artifact id and share URLs stay stable, like a voice recast. Durable media is invalidated; update the recording (POST /artifacts/{artifactId}/update-recording, free) to hear the change — only the lines it changed are spoken again. Requires Idempotency-Key. Scope: artifact:publish.