ElevenLabs v3: FAQ

ElevenLabs v3 converts text into natural-sounding speech, with a library of voices and expressive audio tags for emotion and delivery.

1. What is ElevenLabs v3?

It is a text-to-speech model. You provide text and it produces spoken narration or voiceover. v3 is the most natural and expressive option and supports inline audio tags for controlled delivery.

2. How much text can it speak at once?

Up to 5,000 characters per generation.

3. What are audio tags and how do I use them?

Audio tags are inline cues in square brackets that shape emotion and delivery, for example [excited], [whisper], or [laughing]. Place them in the text where you want the shift: [excited] Welcome to our show!. An Enhance option can add expression tags automatically.

4. Which voices are available?

You can choose from Adam, Bella, Chris, Sarah, Alice, Bill, Brian, Charlie, Daniel, Lily, and Matilda. Examples: Bella is warm narration, Adam is firm male, Sarah is professional, Charlie is energetic Australian, Daniel is a British broadcaster, Brian is deep and resonant, Lily is velvety British, and Alice is a clear British educator.

5. How do I control the delivery?

Three settings shape the voice. Stability sets consistency at 0.0 (creative and variable), 0.5 (natural and balanced), or 1.0 (robust and stable); values round to the nearest of these three. Similarity (0.0 to 1.0, default 0.75) controls how close the output stays to the original voice. Style (0.0 to 1.0, default 0.0) controls expressive exaggeration, from neutral to highly expressive.

6. What output formats can I get?

mp3 (standard, high, low, or phone), pcm (cd, standard, or phone), and ulaw (phone).

7. What is it not for?

Singing. For songs and musical vocals, use the music model instead.

At a glance