Strip the vocals from any track to get a clean instrumental — or isolate the acapella. Powered by on-device AI, so your music never leaves your browser.
Open the Stem Splitter
Go to the DefyAudio Studio and select the Split stems tool.
Upload your song
Add any MP3, WAV, FLAC or other audio file. The AI model loads once, then runs fully on your device.
Choose 2-stem mode
Pick Vocals + Instrumental to split the track into an acapella and a karaoke-ready instrumental.
Separate
The on-device AI isolates the vocals from the music — no upload, no server processing.
Download your stems
Grab the instrumental, the vocal track, or both as WAV files.
Isolate or remove
Get the instrumental (vocals removed) or the acapella (vocals isolated) from one pass.
2 or 4 stems
Go further with 4-stem mode: vocals, drums, bass and other instruments.
On-device AI
HTDemucs runs locally in your browser — no cloud, no upload, full privacy.
WAV export
Download studio-quality WAV stems ready for remixing or karaoke.
A vocal remover separates the singing from the backing track so you can create karaoke instrumentals, sample-clean acapellas, or practice tracks. DefyAudio uses HTDemucs — a state-of-the-art AI source-separation model — running entirely on your device, so even copyrighted music stays private.
Because the separation is AI-based rather than a simple channel trick, it works on real stereo mixes where the vocals and instruments overlap — not just tracks where the vocal happens to be centered. The result is a genuinely clean instrumental and a usable isolated vocal.
Yes. Removing vocals is free with a DefyAudio account (5 daily uses on the free tier), with no watermark on the output.
The AI model downloads once, then runs locally in your browser via WebAssembly. Your audio is processed on your own device and never sent to a server.
Yes. 2-stem mode gives you both the instrumental and the isolated vocal track, so you can keep whichever you need.
MP3, WAV, FLAC, OGG, M4A and more. Stems are exported as high-quality WAV files.