Stamina over sparkle
A showy voice is fun for a minute and tiring for an evening. Pick one you would happily hear read a long descriptive passage.
AI voices
Choose from 137 AI voices for audiobooks, 70 standard and 67 premium, across 13 languages. Play any voice on this page, then upload your book: the studio suggests a narrator and a voice for each speaking character, and you can swap any of them before a single line is generated.
EPUB, PDF, DOCX, TXT or MD · up to 30 MB
Read in your browser. Nothing is uploaded until you generate
A selection from the library you cast from. Narrow it to one language, or to a style such as narrator, character or soothing, and press play.
81 voices
Every voice belongs to a model. The model sets the price, the emotional range and whether bracket tags are performed.
| Model in the app | Voices | Credits per 1,000 characters | About per finished hour | Best for |
|---|---|---|---|---|
| Standard (Gemini Flash TTS) | 70 | 9 | $9 to $12 | Most novels; follows directions and inline tags |
| Lite (Gemini Flash-Lite TTS) | The same 70 | 6 | $6 to $8 | Drafts, proof listening and long nonfiction |
| Premium (ElevenLabs Turbo 2.5) | 67 | 13 | $13 to $18 | Distinct character voices with a steady read; ignores bracket tags |
| Expressive (ElevenLabs v3) | The same 67 | 34 | $33 to $46 | Emotional peaks, laughs, sighs and whispers |
Hourly figures assume about 55,000 characters per finished hour at plan credit prices. Compare plans on the pricing page.
The narrator reads most of the words in a novel, so it is the one choice listeners notice for ten hours straight.
A showy voice is fun for a minute and tiring for an evening. Pick one you would happily hear read a long descriptive passage.
Thrillers like a brisk read, literary fiction and bedtime stories a slower one. Play the same paragraph in two voices before deciding.
If the hero has a deep male voice, a lighter narrator keeps the two apart. Contrast matters more than any single voice.
A Regency romance and a Texas crime novel call for different accents. Match the narrator to the world of the book.
The studio reads gender cues in the text when it proposes a voice. Check older and younger characters by ear and recast where needed.
Voices cover 13 languages, with English the most polished. Cast voices from the language your book is written in.

Same cast in every bookWhen you upload a book, speaking characters are found from dialogue tags such as “said Ivy” or “the captain replied”. The AI audiobook generator then proposes a voice for each one: a narrator for everything outside quotation marks, up to ten named characters, and an “Other voices” role for minor speakers.
Open any role to hear alternatives and swap the voice in a click. Standard and Lite share one set of voices, and Premium and Expressive share the other, so pick the model first and cast within it. Once the cast sounds right, reuse it for every book in a series, and see the multi-voice audiobook page for how lines are split between speakers.
Starting points, not rules. Every role can be recast after you hear a chapter.
| Genre | Narrator to try | Character voices to look for |
|---|---|---|
| Fantasy | Warm, unhurried storyteller | A gravelly mentor, a bright young lead, a cold villain |
| Romance | Intimate and clear | Two leads with strong contrast in pitch and energy |
| Thriller | Tight and low | A calm detective, a nervous witness, a measured antagonist |
| Sci-fi | Crisp and neutral | A flat ship AI, a tired captain, an excitable engineer |
| Children’s | Gentle and expressive | Playful, high-energy voices for animals and sidekicks |
| Nonfiction | Conversational and steady | Usually none; one voice keeps the focus on ideas |
More casting notes in fantasy audiobook narration and romance audiobook narration.
Test on your own text. Previews are written to flatter a voice. Your chapter, with its names, sentence lengths and dialogue, is the real test, so generate a chapter with the suggested cast before you change anything.
Listen at normal speed. Ten seconds tells you whether you like a voice; twenty minutes tells you whether you can live with it. Listen to a full chapter at 1x with the text in front of you.
Direct, don’t recast, for a single line. If a voice reads one moment too flatly, a bracketed note such as [shaken] or [sarcastic] usually fixes it. The AI voice emotion tags page lists the notes that work best.
Know what is not offered yet. You cannot clone your own voice today; voice cloning is on the roadmap. Character techniques from human narration still apply, and how to do character voices is a useful read when you cast.
Check store rules on AI narration. You own the audio you generate from text you hold the rights to, but stores set their own rules for AI voices and many ask you to disclose them. Check each store’s current policy before you publish.
It depends on the book. Standard voices suit most novels at a low cost, Premium voices give character-heavy fiction more distinct personalities, and Expressive voices handle laughter, whispers and grief best. Generate the same chapter twice and compare before you commit.
Playing voices and casting a book are free. Generating audio uses credits charged per 1,000 characters, and new accounts created with Google get 100 gift credits (valid for 7 days, about 12 minutes on Standard; usage varies by text and voice tier) to voice a short sample with any of the 137 voices, no card needed.
Yes. Once a book is cast, reuse the narrator and character voices for the next one, so a recurring character sounds the same in every installment. Stay on the same voice model so the voices are available.
Give them voices that differ in more than one way: gender, age, pitch or pace. Play the two side by side in the studio, and if they still blur in a fast exchange, recast the less important one.
Premium voices on Turbo 2.5 ignore bracket tags. Switch the project to Expressive (ElevenLabs v3) to keep the same premium voices with full tag support, or use Standard or Lite voices, which also perform directions and inline tags.
Yes. The library covers 13 languages, including Spanish, German, French, Portuguese, Italian, Japanese and Chinese. English voices are the most polished, so preview a chapter in your language before generating the whole book.

Upload a chapter, keep the suggested cast or swap any voice, and hear the result with free credits when you sign up with Google.