Beyond Presence adds a real-time video avatar to a conversational AI agent, so a voice bot appears as an animated, lip-synced human face during a call. The company trains its own avatar models (the current generation is Genesis 2.0) and reports 250 ms speech-to-video latency with about one second end-to-end response time.
There are two ways to use it. The speech-to-video path takes your existing voice agent, built with LiveKit Agents or Pipecat, and renders an avatar into the LiveKit room by creating an agent session with an avatar ID and LiveKit credentials. The managed-agents path is no-code: you define an agent in the Studio with an LLM, a knowledge base, and an avatar, and the platform hosts the whole conversation, emitting webhook events and storing transcripts. Both are exposed through a REST API with Python and TypeScript SDKs and a published OpenAPI specification.
Features
- Speech-to-video API: attaches a real-time avatar to a LiveKit session driven by your own voice agent
- Managed agents: create, update, and run hosted conversational video agents without code via the Studio or API
- Avatars: stock avatars available immediately, plus custom avatars exclusive to your account
- Knowledge files: upload PDFs to give an agent document context
- External LLM configurations: connect OpenAI-compatible LLM providers to a managed agent
- Agent behaviors: just-in-time context injection, webcam vision, and wake-word activation modes
- Calls API: list calls and retrieve their transcribed messages; webhooks deliver conversation events in real time
- Embedding: deliver agents through an iframe or the LiveKit client SDK on web and mobile
- Framework support: LiveKit Agents, Pipecat, and n8n, with Python and TypeScript SDKs
- Latency: 250 ms speech-to-video and roughly 1 s end-to-end response
