
Any backend, embodied
Chat Completions, the Responses API, or a push WebSocket. Emotion, motion, voice tag and caption arrive as generate_express cues beside the reply, so she reacts instead of just answering.
A VRM character who lives on your desktop, watches the cursor, and speaks when the agent you connect has something to say.
She perches on a window top, peeks around a side edge, strolls the floor, and jumps between windows. Dropped in mid-air, she falls to the first surface below and lands.

Chat Completions, the Responses API, or a push WebSocket. Emotion, motion, voice tag and caption arrive as generate_express cues beside the reply, so she reacts instead of just answering.

Screen changes, quiet stretches and set times fire cues to the backend. Whether to say anything is the backend's call; YUI only renders what comes back.
YUI ships no model. It renders, senses and fires; judgment, persona and memory belong to whatever you plug in, so she is exactly as smart and as yours as that backend.
One required piece, and it ships in the repo. Add a brain and a voice when you are ready.
Ready to wire her up?
Get started