Audiobook Studio 2.0 docsVoices, Variants & Samples
A voice in Studio is a small hierarchy of its own. Understanding it makes the Voice Lab straightforward.
At a glance
- A Voice is a reusable identity.
- A Variant is a tuned version of that voice for an engine.
- Samples are the audio a variant is built from.
- Voice identity is kept separate from engine assets.
Voice vs variant
A voice is the identity; a variant is a usable build of it. One voice can have several variants, for different engines, speeds, or styles.
Samples & rebuilds
Variants are built from sample audio. Add or replace samples and rebuild to improve a clone. See Voice Cloning Quality.
Engine-per-voice & portability
A voice's identity (name, image, tags) is separate from its engine-specific assets. That's what lets a voice move between engines, and what makes voices portable as shareable bundles.
Coming soon: voices become shareable bundles with an icon, a playable sample, and rich tags, published to a Hugging Face library. See Voice Icons & Tags.