# MOHOLO — Motion Holograms

> MOHOLO builds photoreal live 3D avatars ("gaussian holograms"). One short
> phone video is enough to capture a person as a fully three-dimensional
> talking head that speaks, blinks, holds eye contact and turns in real time —
> rendered by the viewer's own device (WebGL in the browser, or embedded in
> apps and game engines), not streamed as video.

Key facts:
- Input: a single phone video of the subject.
- Output: a rigged, photoreal 3D head driven live by conversation
  (speech-to-face driving, personalized to the subject's own talking style).
- Rendering is client-side: the server streams only motion data (a few kB/s),
  making per-viewer cost roughly 1000x lower than streamed GPU-rendered video.
- Use cases: personal digital twins, and photoreal faces ("embodiment") for
  AI assistants and agents.
- Status: private beta at https://moholo.co/ — live demo avatar (Ana) on the
  landing page; beta signup on the page.

Planned: an embeddable widget so any website can host a MOHOLO avatar with a
custom system prompt, using the site owner's own Gemini Live API credentials.

## LiveKit and Pipecat plugins
Agents on LiveKit Agents or Pipecat add the avatar as a configuration toggle: server side `pip install livekit-plugins-moholo` (`moholo.AvatarSession(avatar="ana").start(session, room)`) or `pip install pipecat-moholo` (`MoholoAvatar(avatar="ana")` in the pipeline); browser side `npm i @moholo/livekit` / `@moholo/livekit-react` / `@moholo/pipecat`. The agent publishes audio only; the visitor's device renders and lip-syncs the head. Source, examples and the server-to-browser contract: https://github.com/moholo-founder/moholo-plugins. Details: https://moholo.ai/embed-docs.html#livekit-pipecat