AI Audiobook Production: Manuscript to Release Workflow

A 100,000-word novel takes a narrator about two weeks in a booth. The AI workflow compresses that to an afternoon — if the workflow is right.

Step 1: prepare the manuscript

• Separate paragraphs with blank lines — long-form mode splits on them, and paragraph boundaries are the most natural cue points
• Strip headers, footers and chapter numbering that shouldn't be read aloud
• Spell out numbers and abbreviations the way you want them spoken

Step 2: pick one voice, keep it for the whole book

Switching voices mid-book is the cardinal sin. Two sources:

• Narrative voices in the community library (filter by narrative_story)
• Clone your own voice or a licensed narrator's (1–3 minutes of clean audio)

Generate one paragraph first to lock in voice, stability and speed, then batch.

Step 3: generate in long-form mode

Studio → Long-form: up to 50,000 characters per job, auto-chunked by paragraph, each chunk generated with its neighbors as context so intonation carries over, merged into one file with a full subtitle track. One chapter per job; the page shows chunk count and credits before you commit.

Step 4: check continuity

Listen at three spots: chapter openings (does the tone "restart"?), the middle of long paragraphs (does it speed up?), dense dialogue (does it get muddled?). Regenerate only the chunk that's off — that's the fundamental advantage over recording.

Step 5: export and publish

• Audio: MP3, one file per chapter
• Subtitles: SRT, for read-along editions or platform captions
• Chapters: organize by file order

Platform note: Spotify accepts AI-narrated audiobooks; Audible's standard ACX channel still requires human narration, so AI titles go through other aggregators with AI disclosure. Check each platform's current rules before release.