VoiceLab

Local Apple Silicon voice studio for Qwen3-TTS models, with a FastAPI backend and React web UI.

Structure

• backend/: FastAPI API, SQLite persistence, MLX inference adapter, tests
• frontend/: Vite + React + TypeScript studio UI
• data/: generated at runtime for voice profiles, renders, and cache
• models/: local MLX model downloads

Backend setup

cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reload --host 127.0.0.1 --port 8000

Frontend setup

cd frontend
npm install
npm run dev

The frontend runs on http://127.0.0.1:5173 and expects the API on http://127.0.0.1:8000.

One-command production run

./scripts/deploy.sh

If you want to run it from Nushell explicitly:

nu ./scripts/deploy.nu

That command will:

• ensure the backend virtualenv and frontend dependencies exist
• run frontend and backend tests
• build frontend/dist
• start FastAPI on http://127.0.0.1:8000

In this mode the backend serves the built frontend, so the app runs from a single URL. For a faster restart during local iteration, use ./scripts/deploy.sh start --skip-tests.