Member Junction
    Preparing search index...

    Video configuration: whether the realtime session carries a synced VIDEO track (a talking-head / avatar out, and the camera in) alongside audio. Absent / enabled !== true ⇒ audio-only (today's behavior). The video track reuses the entire realtime contract — this just opts a co-agent into it and names the avatar to use.

    interface RealtimeVideoConfig {
        avatarId?: string;
        enabled?: boolean;
        provider?: string;
        providers?: Record<string, JSONObjectLike>;
    }
    Index

    Properties

    avatarId?: string

    Provider-specific avatar/character identifier (e.g. a Runway preset or custom avatar id). OPAQUE — passed through to the matching video driver (its shape is a private pact with that driver).

    enabled?: boolean

    Whether the session should carry video. Default (absent / non-boolean): false. When true, resolution prefers a video-capable model and the client captures the camera + renders the model/avatar video; degrades to audio-only when no video-capable model resolves.

    provider?: string

    Preferred video model/avatar provider — an MJ: AI Models Name OR ID of a video-capable realtime model (e.g. a Runway avatar). Optional; absent ⇒ the default video-capable model is resolved.

    providers?: Record<string, JSONObjectLike>

    Per-provider native video settings keyed by provider, merged into the matching driver's open config bag — an OPAQUE private pact with that driver (mirrors RealtimeVoiceConfig.providers).