# Reference Voice Conversion (HF Space)

Convert any vocal to match a **reference voice** (tone color) with [OpenVoice v2] and optional vocal separation via [Demucs]. Built with Gradio for fast UX.

## ✨ Features
- Upload **Reference** (clean 5–20 seconds) and **Track** or **Acapella**
- Optional **Demucs** stem separation to extract vocals from full mix
- Control **style strength**, **pitch**, **formant tilt**
- **Remix** converted vocal with instrumental and gain controls

## 🚀 Deploy
1. Create a new **Hugging Face Space** (Python + Gradio). Hardware: **GPU recommended**.
2. Add these files (`app.py`, `requirements.txt`, `README.md`).
3. (Optional) Add `spaces.yml`:
  ```yaml
  sdk: gradio
  python_version: 3.10
  resources:
    accelerators: ["gpu"]
Commit & run. First build downloads models (~hundreds MB).
🧪 Tips
Reference should be clean, dry (no heavy FX), mono is fine.
Better results