YouTube Audio Separator

FastAPI service that separates YouTube audio into vocals and instrumental tracks.

Setup

# Install dependencies
uv sync

# Set required environment variables
export CLOUDFLARE_ACCOUNT_ID=your_account_id
export R2_ACCESS_KEY_ID=your_access_key
export R2_SECRET_ACCESS_KEY=your_secret_key
export R2_PUBLIC_DOMAIN=your_domain
export API_SECRET_KEY=your_api_secret_key
export REDIS_URL=rediss://username:password@your-upstash-redis.com:6379

# Start services
docker-compose up --scale worker=2
# OR manually: uv run python main.py && uv run python worker.py

Usage

# Start job
curl -X POST http://localhost:5500/separate-audio \
  -H "Content-Type: application/json" \
  -H "X-API-Key: your_api_secret_key" \
  -d '{"search_query": "Song Title Artist"}'

# Check status
curl -H "X-API-Key: your_api_secret_key" \
  http://localhost:5500/job/your-track-id

API

Endpoint	Description
POST /separate-audio	Start audio separation job
GET /job/{track_id}	Get job status and results
GET /health	Health check

Request

curl -X POST http://localhost:5500/separate-audio \
  -H "Content-Type: application/json" \
  -H "X-API-Key: your_api_secret_key" \
  -d '{"search_query": "Song Title Artist"}'

Response

{
  "track_id": "uuid",
  "status": "processing|completed|failed"
}

Configuration

Required environment variables:

Variable	Description
API_SECRET_KEY	API authentication key
CLOUDFLARE_ACCOUNT_ID	Cloudflare account ID
R2_ACCESS_KEY_ID	R2 storage access key
R2_SECRET_ACCESS_KEY	R2 storage secret key
R2_PUBLIC_DOMAIN	R2 bucket domain
REDIS_URL	Redis connection URL

Optional:

• WEBHOOK_URL - Job completion notifications
• CLOUDFLARE_API_TOKEN - Rate limiting
• PORT - Server port (default: 5500)

Deployment

Docker

docker-compose up --scale worker=2