A voice agent you can see listening.
Speak naturally. The agent reads your tone and expression as you talk, then adapts what it says next — in real time.
Private loopCamera frames are cropped in browser; raw frames are not stored.
1
Allow mic access (camera too, if using Face + voice)
2
Say something like "My payment failed and I'm worried" or "I can't find a setting"
3
Watch the fused affect state and the agent's reply adapt
Ready
Realtime stateReady for a conversation
Allow microphone access when prompted. You can interrupt the agent naturally.
Want this in your product?The same detect → update → adapt loop is available as an API and SDK.