Sign in before starting. Upload one supported audio file, wait for the duration check, review the estimated credits, and submit the task. You can keep the original file available for comparison while the vocal extraction runs.
AI Vocal Remover and Acapella Extractor
An AI vocal remover that splits a finished mix into an estimated vocal and an instrumental with the vocals reduced. MP3, WAV, M4A and OGG, up to 50 MB.
Separate a finished mix into estimated vocals and an instrumental with the vocals reduced. Both come back as separate files you can preview before downloading.
01
How to Use the Acapella Extractor
The duration check happens in your browser before anything is sent, which is why the credit estimate appears immediately rather than after an upload completes. If a file is too long, you find out at that point instead of partway through a transfer.
Separation runs on a server and takes roughly as long as the track itself on a typical queue. The page keeps its state while it works, so you can leave the tab open and come back rather than watching a progress bar.
One file goes in per task. If you have a set to get through, run them one after another rather than looking for a batch field — the history list keeps every finished task, so you can collect the downloads at the end instead of waiting at each one.
02
Previewing and Downloading Both Results
When processing finishes, preview and download the estimated vocals and the instrumental with vocals reduced. Both are produced from the same pass, so you never need to run the track twice to get the other half.
Listen to both before downloading, and listen to the quiet parts rather than the chorus. A separation that sounds clean under a loud full-band section often reveals its artifacts in an exposed intro or a final held note, where there is nothing to cover them.
The word estimated is doing real work in those labels. What comes back is a model's best reconstruction of two layers that were mixed together and cannot be perfectly undone, not the original multitrack stems.
03
What People Do With the Two Files
Use the vocals for practice or editing and the instrumental for accompaniment listening or arrangement study. Both are estimates from a finished mix.
Singers take the instrumental to rehearse against the arrangement they will actually perform with, which is closer than a generic backing track. Players do the reverse and keep the vocal to check where phrases sit against a part they are learning.
People studying a production use both halves together: hearing the instrumental alone makes the arrangement audible in a way the full mix hides, because the vocal normally occupies the midrange that everything else is ducked around.
04
Formats, Credits and Separation Limits
MP3, WAV, M4A, and OGG uploads are supported. Files must be under 50 MB and signed-in tasks can be up to 10 minutes. Each started five-minute interval costs 5 credits. Source mixing, reverb, compression, and overlapping frequencies can leave bleed or artifacts, so review important passages.
Credits round up per interval rather than per second, which has one practical consequence worth knowing: a six-minute track costs the same ten credits as a ten-minute one. If you only need a verse, trimming the file first with the audio cutter drops it into the lower bracket.
Material with a dry, centred lead vocal separates best. Heavy reverb puts vocal energy everywhere in the stereo field, and a vocal doubled or layered with harmonies gives the model several things to call the same part, so both tend to leave more bleed than a plain lead.
05
Rights and What You Can Do With the Output
Separating a recording does not change who owns it. An extracted vocal or instrumental carries the same rights as the track it came from, so commercial use needs permission for the source recording regardless of how the file was produced.
That matters most for the two most common uses. A cover performed over an extracted instrumental is still a use of the original recording, and a vocal lifted out of a commercial release stays a part of that release even once it is a separate file.
For your own material the question does not arise, which is the usual case here: most people separating a track are working on something they made, something they have a licence for, or something they are studying privately rather than publishing.
06
Vocal Remover Questions
Can I use an extracted vocal commercially?
Only when you have the necessary rights to the source recording and the intended use. Processing copyrighted audio does not grant commercial permission.
What does a new task produce?
A new task returns estimated vocals and an instrumental with the vocals reduced. Both come from the same pass, so you get both halves from one request.
How are credits calculated?
The task uses 5 credits for each started five-minute interval, with the estimate shown before processing. A six-minute file therefore costs the same as a ten-minute one.
Which formats and sizes are accepted?
MP3, WAV, M4A and OGG, under 50 MB, and up to ten minutes long for a signed-in task. The length is checked in your browser before anything uploads.
Why can I still hear the singer faintly in the instrumental?
Reverb and effects spread vocal energy across the mix, and what remains after the voice is pulled out includes the room it was sung in. Dry, centred vocals leave the least behind.
Are these the original studio stems?
No. They are reconstructions of two layers from a finished mix. Real stems exist only before the mix is made and cannot be recovered from the final file.
Can I get drums and bass separately instead?
Not from this tool, which splits a mix in two. The stem splitter on this site separates a track into four parts when you need the rhythm section apart from everything else.
Can I separate a song I generated on this site?
Yes. Eligible tracks from your generation history can be sent straight through without re-uploading them, which also skips the file size and format checks since the audio is already here.