# live-wizard
24/7 live TikTok wizard. ponytail build: thin glue + best-in-class OSS + Dockerfiles.

## run it
```bash
cp persona/voice-sample.wav persona/avatar.png .env # optional
docker compose up brain tts avatar-cpu router
```

## persona
Edit `persona/persona.yaml` to change the wizard character without changing code.

## services
- `brain`: Ollama + model
- `tts`: Coqui XTTS / F5-TTS with cloned voice
- `avatar-cpu`: Wav2Lip on CPU for prototypes
- `router`: audio loop + RTMP + TikTok Live connector