Source Separation

AI-powered stem extraction with ONNX Runtime

Demucs

Spleeter

4-Stem

ONNX Runtime

Model Source Configuration

Choose where to load ONNX models from

Model Source

Auto Mode

R2 Not Configured

Auto Mode: Tries R2 bucket first for faster loading, falls back to CDN proxy if unavailable.
R2 Bucket: Fast, reliable model loading from Cloudflare R2 storage (requires configuration).
CDN Proxy: Uses proxy to bypass CORS restrictions when loading from external CDNs.

R2 bucket is not configured. Models will load from CDN proxy instead. To enable R2, configure public access for the jewelmusic-onnx-models bucket.

Drop your audio file here

or click to browse

Supported formats: MP3, WAV, FLAC, OGG, M4A

Maximum file size: 100 MiB (105 MB)

Model Settings

Separation Model

• vocals, drums, bass, other

• Sample Rate: 44100 Hz

No Audio Loaded

Upload an audio file to separate it into individual stems

Vocals

Drums

Bass

Other