Generating Audio
Once the cast is approved, starting generation kicks off a pre-flight check before anything renders: any speaking character on the Qwen engine that hasn't had a voice designed yet is flagged, so you can design it now or knowingly proceed with a generic fallback voice.
English books get a "Proceed anyway — generic Kokoro fallback voices" link alongside "Design full cast"; a non-English book has no proceed affordance at all — every speaking character needs a designed voice before it can generate.
Once you're past the gate (and, for a Qwen book, past the "Choose the voice
model" tier prompt), each chapter renders line by line. The Generate screen
shows the whole run's overall progress, an ETA, and a live "Synthesising
<character> · line N of M" counter for the chapter currently in flight.
A finished chapter turns green with a "Done" badge and duration, and picks up Preview / Exclude / Rename / Re-analyse / Regenerate actions. The rest of the book keeps its place in the queue underneath.
Next: The Quality Gate.