Isolate the vocals from a song — the acapella — for sampling, remixing or practice. You get the vocals as the main result, and the instrumental too.
This tool uses AI to separate a song into two parts and hands you the vocals on their own — the acapella — as the main result. The instrumental comes out of the same pass and is there too if you want it.
It runs entirely in your browser. The first time, it downloads a small AI model (about 66 MB) and then does the separation on your own device, using your GPU where available. Your song is never uploaded.
Good enough to sample, study or sing against, not a studio master. On most songs the vocal comes out clearly, though busy or heavily-produced mixes can leave a faint trace of the backing music in the acapella. It is a two-stem split (vocals vs. everything else), so it does not isolate individual instruments.
If you need cleaner isolation, or the individual drums / bass / instrument stems, a higher-quality AI stem separator — a heavier model that runs on our servers and produces four separate stems — is coming soon.