Split a finished song into stems such as vocals, drums and bass, then mute, solo or rework each part on its own.
Preview your file first. When you press start, it is uploaded privately and an AI model separates it on our servers, which can take a few minutes. The price depends on the length of the audio and is shown on the button before you start; if separation fails, the credits are refunded automatically. Play and download each track here, or find them later in your generation history.
Up to 50 MB / 10 minutes per file · 20 credits per minute (rounded up, 2-minute minimum)
or drop it here
MP3, WAV, M4A, FLAC and other common audio formatsYou get four tracks: vocals, drums, bass and other instruments. Guitar, piano, synths and everything else go into the other instruments track.
Your file is uploaded only when you press start. Only your account can open it, and it is shared only with the AI model provider that separates it.
Pick a track you have the rights to and preview it here.
Sign in and split vocals, drums, bass and other parts.
Save each part as its own audio file.
A stem is one group of sounds from a mix, such as all the vocals or the whole drum kit. An AI stem splitter estimates each group from the finished stereo file.
A vocal remover gives you two files: the voice and everything else. A stem splitter also pulls the backing apart, so parts such as drums and bass get files of their own.
Separation models need serious computing power. When you start a split, the file is processed on our server and an AI model provider: upload a song, download stems.
Once a song is split into parts, you can change one part without touching the rest of the mix.
Rebuild a song around its vocal stem, swap in your own drums, or loop a bass phrase as the seed of a new idea. Clear the rights before release.
Lay the acapella of one track over the instrumental of another, or cut a drums-only intro so the next song in your set mixes in cleanly.
Mute the bass stem and play the bass line yourself, or mute the drums and keep time on your own kit with the rest of the band intact.
Lift a clean drum break, a piano chord or a vocal phrase out of a dense mix, instead of sampling the whole song with everything on top of it.
Solo the drum stem of a song you admire to hear how loud the kick sits, how wide the cymbals spread and how the bass and vocal share space.
Mute the vocal stem for a backing track, or keep it low as a guide voice in the background while you sing the lead line over the top.
All three give you separate parts of a song. They differ in how many parts, how clean they are and who can get them.
| Compare | AI stem splitter | Vocal remover | Original multitracks |
|---|---|---|---|
| What you get | Several stems, such as drums and bass | Two files: vocals and instrumental | The real tracks from the session |
| Where it comes from | AI estimate from the finished mix | AI estimate from the finished mix | The original recording session |
| Sound quality | Some bleed between parts is normal | Less bleed with only two parts | Clean, no separation artifacts |
| How to get it | Sign in and upload a song | Sign in and upload a song | Released by the artist or label only |
| Rights | The song's copyright still applies | The song's copyright still applies | Set by the stem or remix license |
| Best for | Remixes, practice and DJ edits | Karaoke, covers and backing tracks | Official remixes and pro mixing |
AI separation is only as good as the file you give it. These habits lead to cleaner stems with less bleed.
Use the highest quality version you have, ideally a lossless file rather than a low bitrate MP3 that has already blurred fine detail.
Phone recordings of a live show or audio ripped from a stream add crowd and room noise, and that noise ends up spread across every stem.
Sparse pop, hip hop and electronic tracks tend to split cleanly. Dense rock, orchestral and live recordings leave more bleed.
Big reverb and delay smear a sound across time, so echo tails may land in another stem. Dry, close sounds separate the most cleanly.
Listen to each stem at the busiest part, then layer them back in the audio mixer. If the sum sounds like the song, the split held together.
Planning a remix or mashup? Check the BPM and key of your stems first, so new parts and other songs line up in time and in tune.
What to know about stems, sound quality, rights and your files.
A stem splitter, also called a stem separator or stem extractor, takes a finished song and splits it into separate parts such as vocals, drums and bass. AI models do this by learning what each kind of sound looks like and estimating it from the full mix.
Choosing a song and previewing it on this page is free. AI separation needs an account and uses credits: 20 per started minute of audio with a 2-minute minimum, from a plan or credit pack, and failed runs are refunded automatically.
No stem splitter is truly lossless. Each stem is the model's estimate of one part, so faint bleed from other instruments and small artifacts are normal. You keep the most detail by uploading a lossless or high bitrate source instead of a heavily compressed copy.
It splits a song into four stems: vocals, drums, bass and other instruments, and each one is ready to play and download. If you only need vocals and an instrumental, the vocal remover is the simpler choice.
A vocal remover makes a two-way split: the vocals and the instrumental. A stem separator goes further and also divides the instrumental into parts such as drums and bass. Pick the vocal remover for karaoke, and stems for remixing, DJ edits and practice.
Split the song into stems, leave the drum stem out and mix the rest back together, for example in the audio mixer. The drum remover does both steps for you and hands back a drumless track plus the isolated drums. Our guide on removing drums from a song covers the full process.
Splitting a song does not change its copyright: the stems belong to whoever owns the original recording. Personal practice is one thing, but releasing a remix or using a sample publicly needs permission. Use music you own, or check the license first.
AI separation does not run in your browser. When you start a split, your file is uploaded to our server and an AI model provider for processing, and it is used only to handle your request. Our privacy policy explains how uploaded files are treated.
Choose a track, sign in and split it into vocals, drums, bass and other stems.
