Upload the song
Drop an MP3, WAV, M4A, AAC, OGG, Opus, or FLAC file. No account is needed for the preview.
To remove vocals from a song, upload it here: an AI model takes out the lead and backing vocals and keeps the music. Hear 30 seconds free, then download the song without vocals as a WAV. Nothing to install.
Last updated
Play the free 30-second preview and switch between the song and the version without vocals. A faint ghost of the voice is normal in songs with heavy reverb; the music should not get duller.
The song as you uploaded it.
The same 30 seconds with the singing taken out.
Before separation models, removing vocals meant tricks that only worked on some songs.
Flipping one side and adding it to the other removes whatever sits in the middle, which also takes out the kick, the bass, and the snare.
Cutting the voice’s frequencies takes the same range out of guitars and keys, and the song sounds hollow.
An old mono recording has no sides to cancel, so the tricks do nothing at all.
Spectral editing by hand works on one line at a time. A model does the whole song in minutes.
Demucs htdemucs, an open-source model from Meta (MIT license), separates the voice from the instruments by how they sound, not by where they sit in the stereo field. It scores 9.00 dB SDR on the MUSDB HQ test set.
Nothing is cancelled out of the center, so the low end stays full. Tools that remove vocals from song mixes by cancelling the center take the kick and bass with them.
Old mono recordings can lose their vocals too, and come back as mono.
You hear 30 seconds of your upload before you pay, not a curated demo.
Most studio songs split cleanly. These are the harder cases.
The echo tail can stay as a faint shadow after each line.
Crowd singing and room sound blur together and leave more of the voice behind.
For a podcast or a vlog with a music bed, use the background music page instead.
Three methods remove vocals from a song. They differ in what else they take out of it.
| Compared on | Center cancelling | EQ cut | Separation model (this page) |
|---|---|---|---|
| How it works | Subtracts one side from the other | Cuts the voice frequencies | A trained model splits the mix by sound |
| Bass and drums | Thinned or gone | Mostly kept | Kept |
| Guitars and keys | Kept if off-center | Hollow in the cut range | Kept |
| Mono songs | Does nothing | Works | Works |
| Where | Audacity, most DAWs | Any EQ | In the browser, no install |
Demucs htdemucs was trained on the MUSDB set plus 800 songs. It splits a mix into drums, bass, other instruments, and vocals; this site keeps the vocals, or adds the other three back together.
Demucs on GitHubYou can remove vocals from song files saved as audio, up to 2 GB. A studio master works best.
The download is a 44.1 kHz WAV. A stereo song stays stereo, and a mono one stays mono. Video files are rejected, so save the soundtrack first.
Save its soundtrack with the free extract tool, which runs in your browser and uploads nothing, then upload that file here.
Drop an MP3, WAV, M4A, AAC, OGG, Opus, or FLAC file. No account is needed for the preview.
Switch between the song with and without vocals. Listen to the loudest part of the vocal.
Process the full song and download a WAV. It is deleted after 24 hours.
Sing or play over the original band.
Use a song’s beat without the lyrics talking over your cues. Check the venue’s music license.
Sing a song live over its own backing.
All the song pages run the same model. Choose by the part you want to keep.
| Job | Use |
|---|---|
| A song with its singing taken out | Use this page |
| Only the vocals from a song | Acapella extractor |
| A sing-along backing track | Karaoke maker |
| Music under a talking voice | Remove background music |
The upload and the result are deleted 24 hours after you upload. They are not public demos.
Monthly and yearly prices stay visible together. You pay after the preview sounds right and you keep the full file. Pro includes 20 hours a month, not unlimited cleaning.
Free
Try it on your own recordings
Upload and preview$9.99/month
$9.99/month, or $95.88 per year ($7.99 per month)
For weekly podcasts and video creators
Get StartedBuy minutes once. They do not expire.
1 Hour Pack
$2.99
Pay once, use whenever you need
5 Hour Pack
$9.99
Lower price per minute for bigger projects
| Plan | Price | Includes |
|---|---|---|
| Free | $0 | Free 30-second previews, Unlimited silence removal, Files deleted after 24 hours |
| Pro monthly | $9.99/month | 20 hours of processing per month, Audio and video, Noise and echo removal |
| Pro yearly | $95.88 per year ($7.99 per month) | 20 hours of processing per month, Audio and video, Noise and echo removal |
| 1 Hour Pack | $2.99 | 60 minutes of processing, Audio and video, Minutes never expire |
| 5 Hour Pack | $9.99 | 300 minutes of processing, Audio and video, Minutes never expire |
Upload the song, play the free 30-second preview with the vocals removed, and download the full song without vocals as a WAV.
The 30-second preview is free and needs no account. The full song uses processing minutes from a Pro plan or a minute pack.
Audacity’s built-in Vocal Reduction and Isolation effect works by cancelling the center of a stereo mix, so it also thins the bass and drums and does nothing on mono songs. A separation model keeps them.
No. All the voices are removed together. For a song where you need the harmonies, keep the original for those parts.
From its soundtrack, yes. Save the audio of a video you have the rights to with the free extract tool, then upload that file here.
No. Each run gives you one WAV: the part this page keeps. To get the other part of the same song, upload it on the acapella extractor page for the vocal. That is a second run, so it uses processing minutes again, and the free preview works there too.
Splitting a song does not change who owns it. Practice, study, and singing at home are one thing. Posting the instrumental or the vocal, or selling a cover or remix built on it, still needs permission from the rights holders.
You can hear the 30-second preview 3 times in 24 hours without an account, and 20 times in 24 hours after you sign in. The upload limit is 2 GB. The full song is billed in whole minutes, rounded up, with a minimum of 1 minute, so a 3:20 song uses 4 minutes.
Remove noise from audio
Wind, fans, traffic, and hum
Remove noise from video
Clean the soundtrack of MP4, MOV, WebM
Remove echo
A repeat you can count after a word
Remove reverb
Shorten a dense room wash
Audio enhancer
Make speech easier to hear
Remove background music
Lower music under someone talking
Instrumental maker
Take the vocal out of a song
Acapella extractor
Keep only the vocals of a song
Karaoke maker
Turn a song into a backing track
Remove silence
Cut pauses, free and in your browser
Extract audio from video
Save the soundtrack as M4A, MP3, or WAV
Normalize audio
Set loudness to -14 or -16 LUFS, free
Increase video volume
Make a quiet video louder, picture untouched
Remove buzz & hum
Lower electrical hum under a voice
Remove static & hiss
Lower hiss and static under a voice
Remove applause & clapping
Lower claps around a speech
Whether you need to clean a podcast, a meeting, or a video you already shot, these guides show the next step.
Upload one song and hear 30 seconds without the singing.
Upload a song