Media-ly Labs · Live Avatar Engine

Sage, procedurally alive.

A real-time 3D head driven by a layered behavior engine — breathing, blinks, eye saccades, gaze-tracking, viseme lip-sync and expression blending, all reacting to app state. Talk to her, switch states, and watch the live drivers on the right.

SAGE

Ambient · breathing

I'm here whenever you need me.

Loading Sage…

Theo

Recognized · Admin

Live drivers

jaw open

0.00

mouth wide

0.00

energy

0.00

blink

open

gaze target

state  : idle
speaking: false
fps    : –

Transcript

Nothing said yet — talk to Sage to start.

Speak responses aloud (browser voice)

Now running on a real rigged model (Ready Player Me). Proper geometry, hair and ARKit blendshapes — the same live behavior engine drives her: jaw/visemes → mouth, gaze → eye + head bones, randomized blinks, and per-state expression blending. Any rigged glTF/VRM (incl. a custom-styled model) drops into this same engine. Voice is browser TTS for the demo; production uses local neural TTS with optional Azure viseme timing.