How to split a song into stems — free, in your browser.

Estimate bass, vocal, drum and other layers with heuristic browser DSP, then preview the bleed and download the useful parts. This is fast local approximation, not AI source isolation.

By: AudioWrench product team · Published: June 13, 2026 · Last technically reviewed: July 15, 2026

What "stems" actually are

A song you stream is a single stereo file: every instrument and voice already glued together into one final mix. Stems are those layers pulled back apart — usually grouped into a handful of parts you can handle separately. AudioWrench's Stem Splitter gives you four: bass, vocals, drums, and other (the catch-all for synths, guitars, keys and pads). Once a track is split into stems, you can build a karaoke version, sample a drum loop, drop the vocal over a new beat, or simply turn one element up or down without re-recording anything.

Producers usually get clean stems straight from the project file. But when all you have is the finished MP3 or WAV, you need to estimate the layers from the mix — and that's exactly what this tool does, right on your device.

How to split a song into stems, step by step

The whole flow lives on one page and takes under a minute:

1. Open the Stem Splitter and load a file up to 200 MB. Input codecs depend on the browser and may include MP3, WAV, FLAC or M4A. The file is decoded locally.
2. Hit "Split stems." The tool runs through four passes — rendering the bass, then the vocals, then the drums via a transient gate, and finally the "other" residual — with a status line telling you which stem it's on.
3. Preview the mix. Press Play mix to hear all four stems playing back in sync. While it runs, tap Solo to isolate one layer, or Mute to drop it out — handy for checking how clean the vocal or the drums came out.
4. Download what you need. Each stem row has its own ⬇ WAV and ⬇ MP3 buttons (MP3 exports at 320 kbps). Grab just the vocal, just the drums, or all four — each file is named after your track so they stay organized.

There is no server processing queue. Speed and practical duration depend on decoded audio size, available browser memory and device performance; long tracks can take longer or fail on constrained devices.

Honest expectations: how clean are the stems?

It's worth being straight about this. The Stem Splitter separates layers using frequency and transient analysis, not a deep-learning model. Bass is the low end of the mix, vocals are pulled from the centre channel in the vocal frequency range, drums come from the high frequencies plus a transient gate that catches percussive hits, and "other" is whatever's left over. That means the parts roughly sum back to the original mix — but you'll hear some bleed between layers, especially between vocals and other instruments that share the same range.

For karaoke-style backing tracks, beat sketches or practice, the estimate may be useful. Dedicated neural separators can produce different results and may run locally or on a server; check each product's data flow. AudioWrench's Vocal Remover also offers instant center-channel DSP plus an optional local UVR neural mode when the lead vocal is the main target.

🎚️ Try the Stem Splitter (free) →

Why split stems in your browser?

The Stem Splitter decodes and processes the selected audio on-device with browser audio APIs, without an AudioWrench processing upload or account. Inputs are capped at 200 MB and decoded PCM can use substantially more memory than the source file.

That privacy-first approach has upsides too: it works offline once the page loads, there's no upload-imposed file-size cap, and you never share copyrighted music with a third party just to take it apart. Free, private, and instant — that's the whole point.

What to do with your stems next

Once you've split a song into stems, the obvious next steps are creative. Solo the drums to build a sampled loop, keep just the vocal to lay over a new instrumental, or mute the vocals to make a backing track. You can also rebalance: drop the imported stems into a DAW or push them through AudioWrench's Mastering Studio to glue a fresh mix together. And if you only ever needed to nudge the low end or carve out a frequency, an Equalizer on the original file might get you there faster than a full split.

FAQ

Keep reading

Related tools