AI Vocal Remover Online
Upload a mix and this vocal remover splits it into an instrumental bed and an isolated vocal stem. Remove vocals from a song for karaoke, or keep only the voice with an acapella extractor — both WAV files are ready to download.
Separation runs as a background GPU job. A few minutes is normal; this tab will update when both stems are ready.
How this vocal remover works
Four steps, no desktop installer. The vocal remover runs on a GPU queue so a three-minute track is not a three-minute freeze in your browser tab.
Upload the mix
Drop an MP3, WAV, M4A, FLAC or OGG file. The vocal remover reads duration in the browser and refuses clips longer than ten minutes on the free tier, so you never wait on a job that will be rejected later.
Queue the split
Your browser uploads straight to object storage, then we release the job onto a vocal remover model that estimates vocals versus accompaniment. Nothing is synthesized; this is source separation on the original mix.
Preview both stems
When the vocal remover queue finishes you get two players. Remove vocals to hear the instrumental karaoke bed. Isolate vocals to audition the acapella extractor output before you download.
Download WAV stems
Keep bg.wav as the instrumental extractor result and vocals.wav as the isolated voice. Version one ships WAV only — no extra transcode step, no surprise bitrate.
Why use this vocal remover
Karaoke maker in the browser
Remove vocals from a song and sing over the leftover instrumental. This karaoke maker does not need a licensed catalog: you bring a mix you have the right to process, and the vocal remover returns a playable bed.
Acapella extractor, not a filter
Isolate vocals for a cover, a mashup, or a harmony stack. The acapella extractor output is a single vocal stem, not a notch EQ pretending to be a vocal remover.
Instrumental extractor for beds
Need a bed for a podcast, a YouTube video, or a rehearsal loop? The instrumental extractor side of this vocal remover is bg.wav — the same stem a karaoke maker wants, ready to loop without the lead vocal sitting on top.
No install, no Python notebook
Other vocal remover tools ask you to run Ultimate Vocal Remover locally, pick a model, and wait on a fan. This vocal remover is a page: sign in, upload, wait on the queue, download. The GPU stays on our side.
What people do with a vocal remover
Practice karaoke at home
A karaoke maker is only useful if the instrumental is the song you actually want to sing. Remove vocals from a song you already own, then loop bg.wav until the chorus sits in your range.
Pull a vocal for a cover
Use the vocal remover as an acapella extractor when you need the original phrasing as a guide track. Isolate vocals, drop vocals.wav under your own take, and mute it before you bounce the cover.
Build an instrumental remix
An instrumental extractor gives producers a bed they can chop. Run the vocal remover once, keep bg.wav, and you are not fighting a lead vocal in every downbeat.
Clean a music bed for speech
Podcasters and course makers often need the groove without the singer. Run the vocal remover, duck the instrumental under speech, and keep the mix legal for the rights you actually have.
Vocal remover FAQ
What is a vocal remover?
A vocal remover is a source-separation tool: it estimates the singing voice in a stereo mix and writes two files — an instrumental bed and an isolated vocal. CosyVoice's vocal remover does that in the browser. You upload a track, the job runs on a GPU queue, and you download bg.wav plus vocals.wav. It is not a simple high-pass filter and it is not a lyric video site; it is a stem splitter aimed at karaoke, covers, and beds.
How do I remove vocals from a song?
To remove vocals from a song, sign in with a verified email, drop an MP3 or WAV up to ten minutes, and start the job. The vocal remover uploads the file, queues separation, and returns an instrumental you can play immediately. That instrumental is the karaoke maker output: sing over it, or drop it under a voiceover. If you wanted the opposite, use Isolate vocals on the same result — one upload feeds both.
Can this karaoke maker isolate vocals as well?
Yes. The same vocal remover job is a karaoke maker and an acapella extractor. Remove vocals gives you the instrumental (bg.wav). Isolate vocals gives you the acapella (vocals.wav). You do not pick a mode before upload; both stems are produced together so you can preview which one you actually need.
Is this an acapella extractor or an instrumental extractor?
Both. An acapella extractor keeps the voice and discards the band. An instrumental extractor keeps the band and discards the voice. This vocal remover writes both stems from one pass, which is why the result panel has two columns instead of a single download. Use vocals.wav when you need an acapella extractor. Use bg.wav when you need an instrumental extractor or a karaoke maker bed.
Which files can I upload to the vocal remover?
WAV, MP3, M4A, FLAC and OGG. Video containers are out of scope for this vocal remover — extract the audio first. Duration is read in your browser and jobs over ten minutes will not start on the free tier. Output is WAV for both stems in version one; we are not transcoding to MP3 yet.
How long does a vocal remover job take?
Minutes, not milliseconds. Separation is a GPU job on a shared queue, so a three-minute song is typically a short wait rather than an instant download. The page polls for status instead of locking your tab. If the queue is busy you will see pending and running states before the two players appear. That is expected for a real vocal remover, not a client-side EQ.
How many vocal remover jobs do I get per day?
Free accounts can run this vocal remover three times per UTC day, with each file at most ten minutes. That is enough to try a karaoke maker pass, an acapella extractor pass, and a spare. When the daily cap is hit the page shows an upgrade CTA instead of starting another GPU job. Pro subscribers are not held to the three-job cap. Sign-in and email verification are required; there is no anonymous vocal remover.