Extract stems from any song for DJ sets and bootleg remixes
DJs and remix producers have wanted stem access forever. Studios sometimes release official stem packs for DJ competitions, but for most of music history the only way to get stems from a random song was to know the engineer. AI 4-stem separation changed that.
What 4-stem gets you
Vocal Remover AI's 4-stem mode returns four files: vocals, drums, bass, and 'other' (which includes synths, guitars, piano, pads, and FX). Each is a full-length file mixed to match the timing of the original — drag them into a DAW and they line up automatically.
For DJs: the canonical use is the a cappella drop. Load the vocals file into Rekordbox/Serato alongside the original track; trigger it isolated over the drums of a different song. Or do the reverse — take just the drum stem from Song A and layer it under the melody stem from Song B.
For remixers: full stem access means you can re-produce any song. Replace the drum programming with your own. Retune the bass. Reverse the 'other' stem for a transition effect. Build an entirely new track around the original vocals.
Quality expectations by genre
Stem separation quality varies meaningfully by source genre:
Rock / pop with acoustic drums — excellent. Drums come out dry and usable, bass is tight, vocals are clean.
Electronic / EDM — drums and bass separate well, but 'other' can be a mess of overlapping synth textures that were never truly distinct in the original production. Expect to do cleanup in the DAW.
Hip-hop with sampled drums — drum separation is nearly perfect. Bass and 808 can be hard to separate from each other since they occupy overlapping frequency ranges; both land in the 'bass' stem.
Classical / orchestral — 'other' is where the entire ensemble lives since drums and bass aren't present in the same role. Less useful for stem extraction in the traditional sense.
Workflow tips
Export at the highest available quality. Pro tier unlocks WAV and FLAC output. For DAW use, WAV is ideal — no re-encoding artifacts when you pitch or time-stretch the stem.
Pre-align in your DAW. All four stems are the same length and timing — drop them onto four tracks starting at bar 1, and they'll play back as the original song. From there you can mute/solo/process at will.
Watch the phase. If you sum the four stems back together, you should hear nearly the original song. If there's phase cancellation, check that all four are at unity gain and starting at the same sample offset.
For Ableton users: stem files play nicely as warped clips. For FL Studio users: drop them in Sampler or a Channel Rack. For Serato Stems: use the separated files as your input, not Serato's built-in (which uses an older model).
FAQ
How does this compare to Serato Stems or Virtual DJ's stem feature?
Those tools separate in real-time during playback using older models. Quality is reasonable for live use but visibly behind Demucs v4 for critical listening. For your DJ pool, pre-process stems with Vocal Remover AI (Demucs v4) ahead of time; for on-the-fly mixing, the DJ software's built-in works.
Can I get a clean 'piano-only' stem for mashups?
Not in our current 4-stem output — piano lands in the 'other' bucket. vocalremover.com offers 5-stem with piano isolated. We're working on adding this.
Try it with 3 free separations.
No credit card required. Your first result is ready in under a minute.