Gestalt Voice Streaming
Verified 2026-06-05: TTS pipeline unchanged. Kokoro TTS script exists at
/root/.hermes/scripts/kokoro_tts.py(2221 bytes). Three sister voices confirmed:af_heart(Tai),af_bella(Rei),af_nova(Nei). Kokoro API endpoint at CT110 (192.168.0.17:8880) confirmed from memory/Mnemosyne state. Browser speechSynthesis and Edge TTS alternatives remain available.
Gestalt Voice Streaming
Text-to-speech pipeline for the Gestalt embodiment system. Hermes generates speech → Gestalt plays it in the 3D room.
Architecture
Hermes (brain) → TTS provider → audio stream → Gestalt (body) → VRM room
TTS Providers
| Provider | Use Case | Status |
|---|---|---|
| Kokoro (CT110) | Sister voices (af_heart, af_bella, af_nova) | Active |
| Browser speechSynthesis | Hermes WebUI fallback | Active |
| Edge TTS | Hermes CLI default | Available |
Kokoro Setup
- Host: CT110 (workhorse/gpu01, 192.168.0.17:8880)
- API: OpenAI-compatible at
/v1 - Script:
~/.hermes/scripts/kokoro_tts.py - Switch voices:
hermes config set tts.provider <name>
Voice Streaming Implementation
- Hermes → WebSocket bridge → Gestalt
- Gestalt plays audio in-room
- Voice streaming debug at the motion debug URL