How to Narrate an Audiobook with Text-to-Speech

Text-to-speech is a good fit for non-fiction, study material, reports, public-domain classics for personal listening, drafts you want to review by ear, and stories for your own family. Dialogue-heavy fiction with many characters and strong emotion is harder: a single synthetic voice cannot act the way a professional narrator does.

• Remove things that should not be read. Page numbers, running headers, footnote markers, image captions and tables of contents usually sound wrong when spoken.
• Rewrite visual elements. Turn tables and bullet lists into sentences. "See Figure 3" means nothing to a listener.
• Expand abbreviations, numbers and symbols the way they should be said. Our script-writing guide has a full checklist.
• Keep a pronunciation list of character names and unusual words with the spelling that makes the voice say them correctly, and apply it everywhere.

A voice that sounds exciting for 20 seconds can become tiring after 20 minutes. Test your shortlist with at least two or three minutes of real text, ideally a passage with dialogue and a passage with description. Many listeners prefer a calm voice at normal or slightly slower speed for books. See how to choose the right AI voice.

Each request in the Studio holds up to 2,000 characters, so a chapter will usually be generated in several parts. Split at paragraph or scene breaks, never in the middle of a sentence. Keep the same voice, speed and pitch for every part and every chapter, and name files clearly (for example chapter-03-part-02) so they can be joined in order.

Use pause tags for scene breaks and chapter openings, for example a [pause:long] after the chapter title. When you join the parts in an audio editor, add a consistent silence between them so the rhythm stays even.

• Listen to every file at least once, ideally while following the text, and note any mispronunciations.
• Fix errors by changing the spelling in the text and regenerating only the affected part.
• Check that the volume is even between parts and chapters; adjust in your audio editor if needed.
• Listen to the joins between parts to make sure no words are cut or doubled.

You need the right to turn a text into audio. That is straightforward for your own writing and for works in the public domain where you live, but not for books you bought or borrowed: owning a copy does not give you audio rights.

If you plan to sell or distribute the audiobook, check the current rules of each store or distributor first. Several of them have specific requirements for AI-narrated audiobooks, such as disclosure or restrictions on which titles are accepted. Our Terms of Use also apply to audio created on this site.