← AI Toolbox

ZEM / AUDIO TOOLS

One Song, Your Chosen Stem

GPU-powered separation. Your selected stems, ready individually as they arrive.

First time separating a song?

Start with your audio

No file selected.

Choose the stems you need

Fine-tuned four-stem model · 10 shifts · 75% overlap · 32-bit float WAV

Up to 10 minutes and 200 MB per upload. Only checked stems are saved and downloaded as WAV files. Vocals arrive first when selected. Allow multiple downloads if your browser asks. Guitar or piano uses the experimental six-stem model. Keep this page open until saving finishes.

Choose a song to get started.

One song at a time. Refresh keeps your current tracks. Starting a new separation or choosing a new file clears the previous song from this browser; downloaded files stay on your device.