What is an acapella extractor?
An acapella is a vocal-only recording without the instrumental accompaniment. In a finished song, the voice has been mixed together with drums, bass, guitars, synths and reverberation. There usually isn't a standalone vocal file embedded inside a normal MP3 or WAV.
An AI acapella extractor estimates which sounds belong to the voice and reconstructs that signal as a separate vocal stem. Song Separator also gives you the other stems, so the same separation lets you preview what was removed or make a karaoke instrumental by muting the vocal.
Individual Vocals WAV versus Download Current Mix
The two export buttons serve different purposes. If you plan to edit the voice in a digital audio workstation (DAW), use the WAV button in the Vocals row. This exports the individual stem without applying the mixer's volume slider. You can change the gain and add effects later in your own software.
If you've adjusted vocal loudness for listening, press S on Vocals, set your desired volume and choose Download Current Mix. That combines the tracks currently audible in the mixer into one full-length 44.1 kHz stereo WAV. When only Vocals is soloed, the export contains the vocal stem at your chosen level. The combined mix is attenuated only if necessary to avoid clipping.
How can you use an isolated vocal?
- Remixes and mashups: place a vocals-only track alongside a new backing arrangement when you have permission to use the recording.
- Lyric transcription: focus on the words and timing with less instrumental interference.
- Vocal technique: listen for phrasing, harmonies, vibrato and breath details.
- Music production: audition EQ or effects and test how a vocal sounds with a new arrangement.
Exporting a vocal stem doesn't transfer rights in the underlying recording or composition. Only use or distribute vocals where you have the required permission.
Why doesn't every extracted vocal sound perfect?
AI separation is an estimate, not access to the original studio vocal master. Lead voices in sparse arrangements may isolate more cleanly than large choirs, stacked harmonies or heavily distorted songs. Some instrumental sounds may bleed into the vocal stem; certain syllables or sustained notes may sound watery or processed. Reverb that was printed into the original song may also remain.
Song Separator is free and runs locally in your browser. Your audio file isn't uploaded to our servers. Both separation models require a capable WebGPU browser and sufficient graphics memory. A current desktop Chrome or Edge is recommended; Safari on Mac needs macOS Tahoe 26 or newer for WebGPU. For crashes or GPU errors, use the troubleshooting guide.