Eleven v3 Character Direction

ElevenLabs' new research preview model, Eleven v3 (alpha), offers a feature that goes beyond simple narration to implement character acting through Audio Tags. Users can use tags like [pirate voice] or [sarcastically] to precisely direct tone, pace, and even a character's personality.

This feature allows changing intonation mid-sentence or having the voice perform specific character Archetypes such as a villain or narrator, all without modifying the script or switching voices. This is ideal for fields where the speaker's individuality matters, such as animation, games, and interactive fiction.

The main types of audio tags are as follows:

• Accents and Dialects: [British accent], [Australian accent], etc.
• Character Types and Roles: [pirate voice], [evil scientist voice], etc.
• Tone and Style: [dramatic], [sarcastically], [whiny], etc.
• Genre Cues: [fantasy narrator], [sci-fi AI voice], etc.

In particular, even in scripts featuring multiple characters, tags can be used to easily switch between voices, allowing a rich Ensemble Cast effect to be achieved with just a single voice track.

However, currently Professional Voice Clones (PVCs) are not fully optimized for v3, which may result in lower quality. To fully utilize v3's capabilities, using an Instant Voice Clone (IVC) or a designed voice is recommended, and PVC optimization is planned for the future.