How to remove vocals from a song free
You can test a song section without creating a profile, installing an app, or entering payment details. The tool uses AI source separation to produce two synchronized files: an instrumental with the vocal reduced and a vocal-focused acapella stem.
- Open the free toolGo to the vocal remover and choose an audio file you own or have permission to process.
- Upload a short sectionMP3, WAV, M4A, and AAC files are accepted. Keep the file at or below 50 MB and two minutes.
- Start the splitThe browser uploads the file for temporary processing. The Demucs model estimates vocal and non-vocal sources.
- Listen to both resultsPreview the instrumental and vocals in the page before using your free download.
- Download the file you needThe first successful split of the day includes both stems. Save them promptly because temporary jobs expire.
What the free allowance includes
| Registration | No account or email required for the core workflow |
|---|---|
| Splits | Normally up to 3 per day per IP address |
| Duration | Up to 2 minutes per upload |
| File size | Up to 50 MB |
| Formats | MP3, WAV, M4A, and AAC |
| Downloads | The first successful split of the day includes both stems |
| Storage | Temporary, scheduled to expire after about 1 hour |
You can preview both stems in the browser. The first successful split of the day includes both downloads. When the free processing budget is full, new free jobs pause until the next UTC day. That pause does not change the free allowance numbers above.
Prepare a song section that is worth testing
Do not upload an entire low-quality recording just because it fits the duration limit. Export the chorus, verse, or passage you need from the best authorized source available. A direct stereo export preserves more useful detail than a phone recording made across a room.
- Choose a section that contains both a busy and a quiet vocal moment.
- Leave a short margin before and after the target phrase so fades are easier.
- Avoid clipping. Distortion creates harmonics that can spread into both stems.
- Use WAV for the cleanest source when available; a good MP3 or M4A is fine for a quick test.
What free vocal removal can and cannot do
The model works from a mixed stereo file, not the original recording session. It often produces a useful karaoke backing track and a clear-enough vocal stem, especially when the lead voice is prominent. It cannot guarantee studio-isolated audio.
Reverb, doubled vocals, cymbals, distorted guitars, and synth leads may overlap with the voice. You may hear small traces in the instrumental or a watery texture around the vocal. Preview the loudest chorus and the quietest ending before relying on the output.
If you need to understand why another technique behaves differently, read the comparison of AI separation and phase cancellation.
Temporary processing and responsible use
No account library is created. The service stores the upload and results long enough to process, preview, and download them, then schedules them for deletion after about one hour. The request IP is used to enforce anonymous limits.
Vocal removal does not change copyright ownership. Use recordings you made, licensed, or have permission to process. A separated stem is still derived from the original song, so public releases, commercial samples, and distributed karaoke tracks may require permission from recording and composition rights holders.