🗣️
Drop a GLB / VRM avatar here
Ready Player Me, Avaturn, Character Studio, or any GLB/VRM with ARKit blendshapes
Or choose a sample avatar from the sidebar
Or choose a sample avatar from the sidebar
-- fps
V2 | ARKit 52-dim
1 Initialize V2 Engine
Student distillation model — 52-dim direct ARKit prediction. Fast, lightweight, perfect for GLB avatars.
2 Load GLB Avatar
Powered by CentrioAI — Take a selfie to create a realistic 3D avatar with ARKit blendshapes for lip sync.
Requires T2 avatar type configured at developer.avaturn.me
MIT licensed — Build VRM avatars from scratch, no account needed. Self-hosted, fully open source.
or choose sample
or load custom
3 Voice Input
Mic streams at 16kHz via AudioWorklet → AnimaSync V2 → ARKit blendshapes → morphTargetInfluences in real-time.
Pipeline:
Mic → AudioWorklet (16kHz PCM)
→ lipsync-wasm-v2.processChunk()
→ Float32Array[52] ARKit values
→ Wolf3D_Head.morphTargetInfluences[]
→ Three.js render loop → Avatar speaks! 🎉
All processing runs locally in your browser.
Mic → AudioWorklet (16kHz PCM)
→ lipsync-wasm-v2.processChunk()
→ Float32Array[52] ARKit values
→ Wolf3D_Head.morphTargetInfluences[]
→ Three.js render loop → Avatar speaks! 🎉
All processing runs locally in your browser.