01○
OpenClaw
Receives conversational context and prepares a structured response.
→ Local prototype · system online
View architecture ↘01 Interactive proof of capability
Context becomes
performance.
Hoshino AI connects conversational intelligence to voice and Live2D expression—turning a typed thought into a responsive, embodied moment.
05expression channels
60motion updates / sec
01coordinated response
Live demonstration
SAFE SHOWCASE MODE✦
Detected expressionlistening
HOSHINO
Ready when you are. Give me a little context.
Voice in this public demo uses your browser. The private working system routes clean dialogue to ElevenLabs.
02 The route
One thought. Four coordinated systems.
The original project runs locally. This site visualizes the exact path without exposing private model files, tokens, or conversations.
02○
Semantic reaction
Maps intent into five continuous channels for face and pose.
→03○
ElevenLabs
Generates spoken dialogue while the controller analyzes its timing.
→04○
Live2D
Synchronizes voice, lip movement, gaze, brows, mouth form, and motion.
Run the demonstration to watch the signal travel
03 What was built
Not an expression switch.
A performance controller.
The system coordinates independent facial parameters with speech timing. Instead of triggering a single preset, it blends emotional signals and returns each channel smoothly to neutral.
01Mouth form
02Brow height
03Brow shape
04Eye smile
05Gaze & motion