AI Stem Splitter: Pull a Song Apart Into Its Parts

A stem splitter answers a simple question: what is actually in this mix? VdoBloom gives you two depths. The shallower one returns the song as two files — the vocal and everything else — which is what a singer or a karaoke render needs. The deeper one returns the parts the song is made of, up to twelve of them: drums, bass, guitar, keys and more, each as a separate downloadable track.

Everything runs from one upload of a song you have the rights to, up to 25 MB and eight minutes. Separation is inference rather than a studio handover — the model is estimating which energy in the mix belongs to which instrument — so the results depend heavily on how the track was recorded and mixed. What comes back is a set of audio files you can load into a DAW, a sampler or a karaoke render.

Split a song into stems

What this tool does

  • VdoBloom's stem splitter has three modes: Vocal remover (vocals plus instrumental, two tracks), All stems (up to 12 stems including drums, bass, guitar and keys) and Karaoke video.
  • The source song can be up to 25 MB and eight minutes long, and every mode requires you to confirm you own the audio or have permission to use it.
  • All stems returns each part as a separate track, so the drums, bass, guitar and keys can be downloaded and used independently.
  • The two-track mode returns the vocal and the instrumental as two files, which is the fastest route to a backing track.
  • Separation quality depends on the mix: a dry, centred lead vocal separates far more cleanly than one buried in reverb and doubled parts.
  • New accounts start with free signup credits, one-time credit packs never expire, and subscriptions are also available.

Choosing between two tracks and twelve

The two-track split is the right answer more often than people expect. If the goal is to sing, to rehearse, or to hear the arrangement without the voice, vocals-plus-instrumental is the faster route and produces no decisions about which of twelve files is the one you wanted. The full stem set is for the cases where you need to touch a single element: replace the drums, study the bassline, sample the guitar.

One practical note about working with many stems: the parts sum back together only approximately. Because separation apportions the same audio energy between files, reassembling all twelve rarely returns the exact original mix. Treat stems as material rather than as a lossless disassembly, and A/B against the original before you build anything on top of them.

Getting the best separation from your upload

The source file decides most of the outcome. A high-bitrate copy of the original master beats a re-encoded download every time, because the model is reading spectral detail and lossy compression removes exactly the fine detail it relies on. If you have both, upload the WAV rather than the MP3.

It also helps to start from the version of the song you actually want split. A radio edit, a remix or a live recording will separate differently from the studio version — a live take brings audience noise and room reverb with it, and the room tone tends to bleed into every stem. For sampling and remixing, the studio master is almost always the better starting point.

How it works

  1. 1.Upload the song and confirm the rights

    MP3, WAV or M4A up to 25 MB and eight minutes. Separation on someone else's copyrighted recording without permission is not something the tool will run.

  2. 2.Pick how deep to go

    Vocal remover for two tracks fast, All stems for the full set of parts. Karaoke video is the third mode if what you actually want is the instrumental with timed lyrics on screen.

  3. 3.Name the song and run the split

    The title is used for the output files, which matters when you are splitting several songs and need to find them again in your library.

  4. 4.Download each part and check it solo

    Listen to the stems one at a time. Bleed between parts is normal — the question that matters is whether the part you need is clean enough for the job you have in mind.

Frequently asked questions

How many stems will I get?

Up to twelve in the All stems mode — drums, bass, guitar, keys and the other parts the model can identify in the mix, each as its own track. In practice the count follows the song: a sparse arrangement returns fewer meaningful stems, and a dense production with layered synths and doubled guitars is where the twelve-part model earns its keep.

Do I get a clean instrumental I can sing over?

Yes — that is the two-track mode, which returns the vocal and the instrumental as separate files. Whether the instrumental is clean enough for a recording depends on the source. A centred, dry lead vocal leaves almost no trace; a heavily reverbed or doubled vocal leaves a faint ghost that becomes audible when you listen to the backing alone.

Can I isolate just the drums to sample them?

That is what the All stems mode is for. Pull the drum stem out, check it in your editor, and use it as a reference or a loop. Do remember that a separated drum part carries the reverb and processing of the original mix, so it will not sound like a dry studio drum take.

What file types and sizes does it accept?

Standard audio files — MP3, WAV, M4A and similar — up to 25 MB and eight minutes long. If your track is longer, trim it first; the form tells you when the upload is over the limit instead of failing partway through the job.

Why does one of my stems sound hollow or watery?

Because separation is an estimate. Where two instruments occupy the same frequency range at the same moment — a bass and a low piano, or a synth pad under a guitar — the model has to choose, and it does not always choose right. Dense, heavily compressed modern mixes are the hardest cases; sparse recordings from a few microphones come apart far more cleanly.

Is it legal to split a song I did not record?

You must own the recording or have permission to use it — the form asks you to confirm that, and it is a real condition rather than a formality. What you then do with the stems, including any use on a platform with content matching, is your responsibility.

Is it free to try, and do I need a subscription?

New accounts start with free signup credits. After that, each split runs on credits: one-time packs never expire, and subscriptions exist for a monthly allowance if that suits you better.

Related tools

Vocal Remover and Karaoke Video Maker

Split a song into vocals and instrumental, then turn it into a karaoke video with lyrics timed to the singing, on five backgrounds, 16:9 or 9:16.

AI Song Cover Generator: Your Song, a Different Record

Upload a song you own and hear it in a new genre, extend it, add vocals to your instrumental or a backing track under your voice. Rights confirmed.

AI Music Generator: A Style Prompt In, a Finished Song Out

Describe a style, paste your own lyrics, pick a model and get a finished song back — instrumental or with vocals. Free signup credits to start.

AI Sound Effect Generator: Describe a Sound, Get the File

Describe a sound in words and get a usable effect back — a one-shot impact, whoosh or door, or a longer ambience bed and seamless loop with tempo and key.

Ready to try it?

Split a song into stems