A companion is defined once. That definition drives every interaction — what it says, how it sounds, and how it looks — consistently across all modalities.
Identity
You provide:
- Name — who the companion is
- Backstory — history, context, relationships; this shapes how the companion talks about itself and responds to personal questions
- Personality — how the companion communicates: tone, values, speech patterns, what it cares about
The personality and backstory are applied to every conversation turn. The companion doesn't need reminding who it is — that context is always present.
Visual Character
You provide a visual description or a reference image. We generate a character skin — a consistent visual identity used across:
- Generated images — images of the companion in conversation context
- Avatar video — the talking head rendered in real-time or on-demand
- Media assets — any visual content featuring the companion
The same character appears everywhere. A generated image, a real-time video session, and an on-demand response clip will all show the same face. Visual consistency is maintained automatically — you don't manage individual assets.
Personas
A companion can have multiple personas — variations on the core identity adapted for different contexts or user segments. Personas share the companion's underlying personality but can differ in name, backstory, and presentation.
Each user can be assigned a default persona, and personas can be switched at runtime.
Related APIs
See the API Reference for companion creation and management endpoints.