Free · No sign-up · 100% in your browser
Stem Splitter: Split Songs into 4 Stems
This free AI stem splitter turns one song into 4 stems: drums, bass, vocals and other instruments. Preview each part and download four separate WAV files. Processing runs on your device, with no upload or account.
Drag & drop a song here
or
MP3 · WAV · FLAC · M4A · OGG
File:
The model download is cached for next time.
Your four stems are ready
No upload · No login · No watermark · Audio stays on your device
4 stems from one song
One mixed track goes in. Four separated stems come out, each one playable and downloadable on its own.
Drums
The whole kit — kick, snare, hats, cymbals and percussion.
Practise along, resample a break, transcribe a groove.
Bass
Bass guitar, upright and synth bass — the low end on its own.
Learn a bass line by ear, rebuild the low end in a remix.
Vocals
Lead and backing vocals, lifted out of the mix.
Make an acapella, study phrasing, build a mashup.
Other instruments
Everything else — guitars, keys, strings, synths and horns.
Isolate a riff, make a backing track, study an arrangement.
Every stem is exported as 16-bit stereo WAV at 44.1 kHz — ready to drop straight into Ableton, Logic, FL Studio, Reaper or any other DAW.
Split stems without uploading your song
Cloud-based tools process uploaded audio on a server. Here, the separation happens on your device, which can be useful when you are working with an unreleased demo.
This one does the AI separation inside your browser. The model is downloaded when you start processing, and your song is split locally — never uploaded, never stored, never seen by anyone.
- No file ever leaves your device
- No server in the loop — so nothing to leak
- No account, no email, no upload queue
- No audio is sent to a server
0 bytes uploaded · 0 bytes stored
How to split a song into stems
Add your song
Drop in an MP3, WAV or FLAC. It is decoded locally — no upload.
Choose and start
Choose a sample or full song, then start. First use downloads 80–180 MB plus runtime files; use Wi-Fi.
Split into stems
The model separates the mix into drums, bass, vocals and other instruments.
Download
Preview each stem, then save the ones you need as WAV.
Who uses a stem splitter
Producers & remixers
Pull a clean drum break, a bass line or a vocal out of a reference track and build something new around it.
Musicians learning by ear
Mute everything but the part you are learning. Hearing the bass on its own beats guessing at it under a full mix.
DJs & mashup makers
Split two songs into stems and layer the vocal of one over the instrumental of another.
Karaoke & backing tracks
Drop the vocal stem for karaoke, or drop the drums for a drumless practice track.
Transcribers
Isolate one instrument, then use Sound to MIDI to turn the clean stem into editable notes without the rest of the arrangement in the way.
Audio engineers
Rebalance a mix you no longer have the session for, or rescue a part from a bounced master.
Need just one stem?
Same engine, same privacy, pointed at a single job. Each of these gives you a finished mix rather than four separate parts.
Stem splitter FAQ
What is a stem splitter?
A stem splitter is a tool that takes one finished, mixed song and separates it back into its individual parts — called stems. This stem splitter uses an AI source-separation model to split any track into four stems: drums, bass, vocals and the remaining instruments. You can play each one and download it as a WAV file.
Is this stem splitter free?
Yes, completely. There is no account, no sign-up, no export limit and no watermark on the stems. There is nothing to install either — it runs in the browser tab you already have open.
Do my songs get uploaded to a server?
No, and this is the main thing that sets it apart from other online stem splitters. The AI runs inside your browser, so your song is decoded and split on your own device. Nothing is uploaded, nothing is stored, and nobody — including us — ever receives your audio.
How many stems do I get?
Four: drums, bass, vocals and instruments (everything else). That is the standard four-stem layout used by the Demucs model, and it covers the overwhelming majority of what people want to split a song into.
How long does it take to split a song into stems?
Speed depends on your device. Eligible devices use local GPU acceleration after a compatibility check; other devices use standard processing, which can take tens of minutes for a full song. First use downloads about 80–180 MB of model data plus runtime files. Your browser may cache these files. Try a 30-second sample first if unsure.
What audio formats does the stem splitter accept?
MP3, WAV, FLAC, M4A and OGG all work. Every stem comes back as a standard 16-bit stereo WAV at 44.1 kHz, so it drops straight into any DAW.
How good is the separation quality?
It uses Demucs. Separation is an estimate: dense mixes, reverb and overlapping instruments can leave bleed or artifacts. Listen to each stem before using it. The other stem groups guitars, piano, strings and similar instruments together; it does not separate each instrument individually.
Can I use the stems in my own music?
Use recordings you own or have permission to process. Separation does not grant rights to the original recording or composition. Check the source license before sharing, sampling or releasing a stem.