THE AI AUDIO LAB

Hear every layer.

Upload a song from your phone, isolate vocals and instruments, and download your stems. No account, no paid AI API, no audio uploads.

Drop in your audio

Choose a track to separate.

Drag & drop your track
or browse files to get started

MP3, WAV, M4A + browser-supported formats · UP TO 90 MB

Choose your separation

How many layers do you want?

VOCALSDRUMSBASSOTHER

Song length

Start small if your phone has limited memory.

Tip: A 30-second test reduces song processing, but the AI still needs around 1.1 GB of memory just to initialize. Newer phones work best.

Export quality

Choose a file type for your stems.

Uncompressed audio · Best for editing and remixing

AI model not loading? Try a saved model file

Automatic downloads use Hugging Face and are free. If loading fails, save the appropriate large AI model to your phone, then choose it below. This can fix download restrictions, but it cannot fix low RAM.

Download free 4-stem ONNX model (166 MB) Choose saved .onnx model from Files

Automatic model download selected

Audio stays on your device. The free AI model downloads once per browser cache (about 136–166 MB). Keep this tab open during processing.