Stem splitter: split any song into 4 stems in your browser

Upload a track and get four separate stems back — vocals, drums, bass and other — using a Demucs v4 model that runs entirely in your browser, with no upload. HD Pack and Pro add server-side HD quality and six stems.

The 4-stem browser separation carries no charge. Server-side HD, 6 stems, batches and FLAC are paid — see pricing.

Split into 4 stems

or drop it here · MP3, WAV, FLAC, M4A, OGG · up to 50 MB / 10 min

Browser separation runs locally · Uploads only if you choose HD

How the 4-stem split works

SplitVocals uses Demucs v4, a hybrid transformer model trained specifically to separate music into four sources: vocals, drums, bass, and everything else (guitars, keys, pads, synths — grouped as "other"). The model analyzes the full mix and reconstructs each source as its own audio stream, rather than just filtering frequencies, so instruments that overlap in pitch still come out as distinct stems.

The whole pipeline — decoding your file, running the model, and re-encoding the output — happens on your own CPU or GPU through WebAssembly or WebGPU. There's no server in the loop, so there's no upload step and no queue.

Who uses a stem splitter

Producers and DJs pull the drum or bass stem from a reference track to build a mashup, or lift an acapella to lay over a new beat. Remix artists use isolated stems as raw material without re-recording anything. Sample-based musicians dig individual instrument lines out of old records for new productions. Because output stems are yours to use in original work you have the rights to, they drop straight into a DAW project.

Getting the cleanest 4-stem split

Feed it the highest-quality source you have — a WAV or FLAC gives the model more detail than a heavily compressed MP3. Dense mixes with lots of reverb or many instruments sharing the same frequency range are the hardest case for any separation model, so expect some bleed between stems (a little vocal breath in the "other" stem, for example) on those tracks. Simpler, well-mixed arrangements split the most cleanly.

Each stem downloads as WAV or 320k MP3, and you can grab all four at once as a ZIP.

Why a browser-based splitter beats an upload queue

Cloud-based splitters typically meter you by the minute and make you wait for a render queue to clear. Running the model on your own hardware removes both limits: no per-minute cost, and processing starts the moment you drop the file. On a GPU-accelerated browser like Chrome or Edge, a full song typically finishes in one to three minutes; Safari and Firefox use a CPU fallback that takes longer but produces the same output.

How it works

  1. 1

    Drop a song

    MP3, WAV, FLAC, M4A, OGG or AAC, up to 50 MB or 10 minutes.

  2. 2

    AI splits it on your device

    A Demucs v4 model runs locally via WebGPU or WebAssembly — nothing is uploaded.

  3. 3

    Play, solo, download

    Preview each stem, then export WAV, 320k MP3, or a ZIP of all four.

Why running locally matters

Private by construction

Your audio is decoded and processed inside the browser tab. There's no upload step, so there's nothing to verify in a privacy policy — check the Network tab yourself.

No queue, no waiting

Processing starts the moment you drop a file. There's no server-side render queue, so a slow day for other users never slows you down.

Works offline after first use

The AI model downloads once (~170 MB) and is cached by your browser. Come back tomorrow, even without a connection, and separation still works.

No per-song cost

There is no per-song server compute to pay for in the browser tier, so nothing is metered. Server-side HD is the paid exception, and it says so before you click.

What is included, and what is paid.

The 4-stem browser separation on this page carries no charge and no cap on songs. The HD Pack and Pro exist for the parts that need a server: studio-grade HD quality, more stems, batches and lossless files.

Browser separation

$0

  • 4-stem separation, on your device, no cap on songs
  • No account, no upload, no queue
  • MP3 and WAV downloads, no watermark
  • Supported by ads

HD Pack and Pro

HD Pack $4.99 one-time · Pro $5.99/month or $39/year

  • Server-side HD vocals and instrumental, cleaner than a browser can manage
  • 6 stems, adding guitar and piano (Pro)
  • Batches of 5 files and lossless FLAC export (Pro)
  • No ads (Pro)
See pricing

Frequently asked questions

How many stems does this stem splitter produce?
How many stems does this stem splitter produce?
Four: vocals, drums, bass, and other (guitars, keys, synths, and anything else in the mix). You can download them individually or all together as a ZIP.
Is the 4-stem split paid?
Is the 4-stem split paid?
No — the 4-stem separation on this page runs on your own device instead of a metered server, so it carries no per-song charge and no daily cap. The paid HD Pack and Pro add server-side HD quality, six stems, batches and FLAC; the pricing page lists exactly what each costs.
What audio quality do the exported stems have?
What audio quality do the exported stems have?
You can export lossless WAV at the source sample rate or MP3 at 320k. Quality is otherwise limited by the source file — a lossy MP3 in gives a lossy-quality separation out.
Can I use the split stems commercially?
Can I use the split stems commercially?
Yes, for audio you own or have the rights to use. You're responsible for the rights to the source material; SplitVocals doesn't add any restriction of its own on top of that.
How long does splitting a song take?
How long does splitting a song take?
Roughly one to three minutes for a typical song on a laptop with WebGPU (Chrome or Edge). Safari and Firefox fall back to a CPU-only path, which takes longer, and phones are slower than desktops either way.
Does the stem splitter work with any music genre?
Does the stem splitter work with any music genre?
Yes, it's genre-agnostic — the model was trained on a broad range of commercial music. Very sparse electronic tracks and very dense wall-of-sound mixes sit at opposite ends of how clean the split comes out, but both are supported.