AI Vocal Remover
Remove the vocals from a song and keep the music. Sign in, upload a track and hear a free 30-second preview with the vocal-level slider. Full-length instrumentals and acapellas come with the Producer plan.
Free 30-second preview when signed in · full tracks cost 2 credits on the Producer plan.
What does an AI vocal remover do?
An AI vocal remover takes a finished, mixed song and estimates two new recordings from it: the music with the singer taken out, and the singer with the music taken out. The instrumental is what most people are after — a karaoke track, a backing track to sing or play over, or lyric-free music for a video. The isolated vocal, often called an acapella, is the same process read the other way.
Older vocal removers worked by phase-cancelling whatever sat in the centre of the stereo image, which is why they left a hollow, watery instrumental and only worked on some mixes. This tool runs Demucs, a neural source-separation model trained on paired mixes and stems, so it predicts what the voice contributed to every moment of the track instead of relying on where it was panned. Reverb tails on the vocal and dense harmonies still leave traces, and no separator returns the original multitrack — but on a typical pop, rock, hip-hop or acoustic mix the result is clean enough to sing over, remix, or study.
Unlike the Stems Separator on this site, which hands you two files and stops, the vocal remover plays both stems together through one slider. Drag it to zero for a pure karaoke version, leave a little voice in as a guide while you learn a part, or push it back to 100% to hear the original balance. Then download whichever file you need.
AI vocal remover features
Instrumental and acapella from one upload
Every job returns two MP3s: the song with vocals removed and the vocal on its own. Download either or both.
Vocal-level slider
Both stems play in sync through a single slider. 0% is karaoke, 100% is the original mix, and anything in between is a guide vocal at the volume you choose.
Demucs htdemucs on a GPU
Separation runs on the hybrid transformer Demucs model — the open model that set the quality bar for source separation — not a phase-cancel trick.
Six input formats, 25 MB, 6 minutes
MP3, WAV, FLAC, AIFF, OGG and M4A are accepted. Files up to 25 MB and 6 minutes long; longer tracks are refused rather than cut short.
Your upload is deleted after the job
The file goes straight from your browser to storage, the model reads it once, and the original upload is removed when separation finishes. Only the two result stems are kept for you.
Free preview, then Producer
Every signed-in account gets three free 30-second previews a day at no credit cost. Full tracks are a Producer-plan feature at 2 credits each, refunded automatically if a job fails.
Who removes vocals from songs
Singers making karaoke and practice tracks
Take the original singer out and keep the arrangement. Leave 20–30% of the vocal in while you learn the melody, then drop it to zero for the performance.
Producers and DJs pulling acapellas
Isolate a vocal for a remix, a mashup or a bootleg edit, or grab the clean instrumental for a set. Check the result against the original with the slider before you commit.
Video and podcast creators
Turn a track with lyrics into lyric-free background music so speech in your video is not fighting a singer. Rights to the original recording still apply.
Music students and teachers
Hear how the arrangement sits under the melody, transcribe a part without the vocal in the way, or build a play-along with the instrument you are studying.
How to remove vocals from a song
1. Sign in and upload the song
Drop an MP3, WAV, FLAC, AIFF, OGG or M4A up to 25 MB and 6 minutes. It uploads straight to storage, not through our web servers.
2. The model separates voice from music
Demucs estimates the vocal, drums, bass and other instruments across the whole track on a GPU, then sums everything except the voice into one instrumental. A full song usually takes a minute or two.
3. Hear the preview, set the vocal level, download
The first 30 seconds come back free. Play both stems through the slider to hear the karaoke version, a guide vocal or the original balance, then run the full track on Producer and download the instrumental, the acapella or both as MP3.
AI vocal remover FAQ
How do I remove vocals from a song?
Upload the track on this page. The separation runs on our GPUs and returns two files, the instrumental and the isolated vocal, in a player with a vocal-level slider. Nothing to install and nothing to configure.
Is this AI vocal remover free?
The preview is. Sign in and any track gets a free 30-second separation, three per day, with the full vocal-level slider and no credits spent. Downloading the complete instrumental and acapella is a Producer-plan feature at 2 credits per track, refunded if a job fails. We would rather say that plainly than hide the limit behind a download button.
Can AI perfectly remove vocals?
Not perfectly. A separator predicts what the voice contributed to the mix; it does not have the original multitrack. Reverb tails and delay throws on the vocal tend to stay with the vocal stem, stacked harmonies can leave a faint trace in the instrumental, and loud cymbals or synth pads sometimes bleed into the vocal. Sparse mixes with one clear lead separate cleanest.
What is the best AI vocal remover?
For most songs the modern neural separators — Demucs, which this tool runs, and the models behind LALAL.AI, PhonicMind and Ultimate Vocal Remover — produce results in the same class, and the differences are per-song rather than per-product. Judge a remover on the mixes you actually need: upload one, listen with the slider at 0%, and check the choruses where harmonies stack.
What formats and limits are supported?
MP3, WAV, FLAC, AIFF, OGG and M4A, up to 25 MB and 6 minutes per file. Longer or larger files are rejected before any credits are spent. Output is two 192 kbps MP3 files.
Can I use the instrumental commercially?
Removing the vocal does not change who owns the recording. Karaoke at home, practice, study and personal remixes are one thing; releasing, monetising or distributing an instrumental or acapella made from someone else's song needs permission from the rights holders, exactly as sampling does. If you own the master or have a licence, the output is yours to use.
What happens to my uploaded file?
It is uploaded directly from your browser to our storage, read once by the separation job, and deleted when the job finishes. The two result stems stay in your library so you can come back to them. We do not train models on your uploads.
Remove the vocals. Keep the song.
Free 30-second preview after sign-in. Full instrumental and acapella on Producer.
More AI music tools
One place to generate, convert, detect, and remix.
Turn a prompt into a full song with vocals.
Write full song lyrics from a theme or a line.
Sung brand hooks and sonic logos with target lengths.
Shape phonk with BPM, key, energy, cowbell and 808 controls.
Build lofi beats by scene, tempo, key, texture and length.
AI rap diss with lyrics + beat in seconds.
Stage name ideas by scene — trap, drill, classic.
Convert an MP3 into editable MIDI notes.
Split a track into vocals and instrumental.
Add a slowed + reverb tape vibe, free in your browser.
Find any track's tempo by tap or upload.
Find whether a song counts in 4 or in 3.
Detect the musical key of any song.
Check your pitch accuracy note by note.
Get key, Camelot code and tempo from one upload.
Batch-key up to 25 tracks into one table.
Find the key through your mic — no file needed.
Get every beat position, bar by bar.
Tap out a tempo, get delay times in ms.
Every note and chord in any key.
Find the note and frequency of any sample.
Live chromatic tuner in your browser.
Turn a melody into note names and timings.
Measure your vocal range low to high.
WAV, FLAC, AIFF, OGG, M4A or MP3 into editable MIDI notes.
Piano audio to engraved sheet music, both hands.
Any recording into a readable, printable score.
AI vs by ear vs hiring — and the fast path.
6 formats in — score, PDF, MusicXML, MIDI out.
Tape warmth, wobble and vinyl crackle on any track.
Guitar, bass, ukulele and more — notes, Hz and a tuner.
BPM to ms for every note value.