How does Remove Vocals isolate instrumentals from a song?

Remove Vocals (remove-vocals.com) separates a mixed song into vocals and instrumental by running the audio through an AI model built on the open-source Spleeter library. You upload an MP3 or WAV, the site processes it on its server, and you download the resulting instrumental from a download page. It is free for up to 2 songs per day and requires no registration.

The separation mechanism

The site describes its service as using artificial intelligence and being "based on the open source library spleeter." Spleeter is a source-separation library: it takes a mixed audio file as input and predicts which parts of the signal belong to different stems (for example, vocals versus accompaniment), then writes those stems out as separate audio files.

In practice, for this tool that means:

• Input: one mixed track containing both vocals and instruments (MP3 or WAV).
• Processing: the AI model estimates the vocal component and subtracts/isolates it, leaving the instrumental.
• Output: an instrumental track you can download and use as a karaoke backing.

This is a prediction task, not a perfect filter. The model works on the mixed waveform, so results depend on how prominent and how "typical" the vocals are in the mix.

What you get, and what "vocals only" means

The main output is the instrumental (karaoke) track. If you want the opposite — the isolated vocals — the site points you to a separate tool, acapella-extractor. So the two use cases are split across two services rather than offered as one combined export.

The site also hosts an Examples section where you can listen to isolated instrumentals before deciding whether the quality is good enough for your purpose. That is the most reliable way to judge output quality, since separation results vary by song.

Limits and access conditions

The length and size caps are stated as necessary to avoid saturating the server, so files beyond 10 minutes or 80MB will not be processed.

How to run a separation

1. Open remove-vocals.com.
2. Drop your audio file onto the upload area, or click to select it. Use MP3 or WAV, up to 80MB and 10 minutes.
3. Wait while the file uploads and the AI processes it — you are redirected to a download page.
4. When processing finishes, click the download button to save the instrumental.

Expected result: a downloadable instrumental version of your track. If nothing downloads, check first that the file is within the size and length limits, since those are the documented hard constraints.

Where quality can fall short

Because the separation is model-based, expect artifacts when:

• Vocals overlap heavily with instruments in the same frequency range.
• The mix is dense or heavily compressed.
• The vocal is quiet relative to the backing.

The site's own Examples page is the practical reference point — compare your track type against those samples before relying on the output for a performance or release.