Characters
One renderer. Fifteen faces.
Every character is a trained model that runs on the device, with a voice of its own. All fifteen below are published today; tap one to hear it.














YOURS
Download sizes are the first-visit transfer; returning visitors load from cache. The realistic characters need a newer web SDK than the one on npm today (0.1.1); ask us for access. Each character passes the same quality battery before it ships.
Custom characters
Your identity, running on your users’ phones.
What we need from you
- 01A reference for the identity: artwork, a design brief, or footage you own the rights to.
- 02Written confirmation you hold those rights. We do not clone real people without documented consent.
- 03A voice direction, if you want the character to sound a particular way.
What comes back
- MODELA quantised renderer for that identity, signed and delivered over the air.
- REPORTThe quality battery it passed: closure, jitter, mouth contrast, held-out reconstruction.
- LICENCESetup fee plus an annual licence. Rendering it then costs nothing per frame, like everything else here.
Turnaround is days, not minutes, and there is no self-serve path yet. That is a real limitation and we would rather you hear it here than in week three. Start a character.