Webnovel Serial Audiobook

Problem

Serial webnovels are updated chapter by chapter, but ad hoc audiobook conversion can produce inconsistent proper-noun pronunciation, drifting narrator tone, and costly rework. This skill targets finalized chapters that must be delivered as serial audio while keeping the same narrative voice across future chapters.

How It Works

It interacts with Beatra through the bundled scripts/mcp_client.py instead of a host connector or REST/OpenAPI fallback. The core flow is:
- Create a chapter ledger with chapter name, order, block labels, and a pronunciation glossary, then normalize the chapter into readable speech text.
- Confirm the voice path: if a narrator sample is provided, verify consent before uploading and freezing the voice_id; otherwise choose a catalog voice through beatra.voices.list.
- Synthesize by block using beatra.speech.synthesize, then poll each task to completion with beatra.tasks.get.
- Re-record only problematic blocks when a phrase is unclear or mispronounced, using a new request instead of rerunning the whole chapter.
Later chapters reuse the same voice_id to preserve continuity.

Boundaries

It is meant for serial webnovel audio, not one-off literary audiobooks, course narration, or personal IP voice-series production. Required inputs include finalized chapter text, chapter order, target language, and a pronunciation glossary for repeated names, terms, or place names. Keep each submission under 50,000 characters and split it into complete sentences or paragraphs, excluding titles, author notes, and bracket tags from synthesis.