🔒 Your files stay on your device — core audio processing runs locally. Privacy details →
Find a toolCtrl K
🎚️
Free · In-browser · No upload

Stem Splitter.

Create approximate bass, vocal, drum and other stems with fast local DSP — no AI model or server queue. Solo or mute them in a live mix, then download each one separately.

Before-and-after illustration of Stem Splitter: Full song mix becomes Downloadable stems through Separate source groups.
Before → process → afterSee the signal story, then hear it in the workspace.

SPECTRAL MAP

Four-Stem Spectrum

See where tone changes across the frequency range. Read the curve from low to high frequency, then connect every visible move to something you can actually hear in the source.

Before-and-after illustration of Stem Splitter: Full song mix becomes Downloadable stems through Separate source groups.
Full song mixSeparate source groupsDownloadable stems
A frequency-domain view: the source profile, the selected spectral move and the resulting tonal balance share one coordinate system.
Open the full signal breakdownBest source, processing logic, limits and three-pass listening check
  1. Full song mix

    Start with the right material

    Practice mutes, rough stem previews and quickly inspecting parts of a conventional stereo mix.

  2. Separate source groups

    Understand the transformation

    Mid/side relationships, frequency bands and transient emphasis are combined into four approximate outputs.

  3. Downloadable stems

    Verify the usable result

    Four locally generated stem estimates with solo, mute and individual download controls.

Stem isolation

Where this workflow stops

Sources overlap in real music, so bleed and artifacts are expected. It will not match dedicated neural stem-separation software on dense or unusual mixes.

THREE-PASS CHECK

Make the final decision by ear

  • Sweep or solo only to locate a problem; make the final decision in context.
  • Watch broad tonal balance as well as narrow resonances.
  • Recheck level after any boost so the comparison stays honest.

How to split a song into stems

  1. Drop a track into the upload panel — any format your browser can decode (MP3, WAV, FLAC, OGG, M4A).
  2. Press the split button and let the separation run on your device.
  3. Audition the result in the live mix: solo or mute Bass, Vocals, Drums and Other to hear each layer on its own.
  4. Download the stems you need — each one exports individually as WAV or MP3.

This stem splitter estimates four layers with local frequency, phase and transient DSP. It is useful for rough karaoke, practice and remix sketches, but overlap between vocals, instruments and drums can create substantial bleed or artifacts. It is not an AI separator and does not promise studio-isolated stems. Inputs are capped at 200 MB; long tracks can take more time or exceed device memory.

Stem Splitter: quick answer and technical limits

Quick answer: A fast heuristic stem splitter that estimates vocals, drums, bass and other content with stereo, frequency and transient cues; it is not an AI separation model.

Best for
Practice mutes, rough stem previews and quickly inspecting parts of a conventional stereo mix.
How it works
Mid/side relationships, frequency bands and transient emphasis are combined into four approximate outputs.
What you get
Four locally generated stem estimates with solo, mute and individual download controls.

Know before you use it: Sources overlap in real music, so bleed and artifacts are expected. It will not match dedicated neural stem-separation software on dense or unusual mixes.

Public browser workflow evidence

Question: Can the real UI export bass, vocals, drums and other WAV stems with the expected authored routing and a sample-level sum close to the input fixture?

Observed result: All four PCM16 WAV stems exported and summed back to the fixture with RMSE 0.00003146 and peak error 0.00009155; authored bass, vocal and transient/side routing checks passed.

Evidence boundary: The splitter is a frequency/transient heuristic, not an AI source-separation model. The fixture verifies deterministic routing and reconstruction, not perceptual isolation on real music.

Network observation: No audio-payload or cross-origin requests were observed during this published workflow.

Open the public benchmark, method and fixture →

Privacy: Selected audio is processed in this browser and is not uploaded to AudioWrench. Normal page assets can still be requested as described in the privacy notice. Privacy details →

FAQ

How does the stem splitter separate a song?
It combines frequency-band analysis with phase separation: bass is captured from the low end, drums from transient energy, vocals from centered mid-range content, and everything left over lands in the Other stem. It is a fast approximation that runs on your device, not a server-side neural model, so expect good — not surgical — separation.
Which stems do I get and in what formats?
Every song is split into four stems: Bass, Vocals, Drums and Other. Each one has its own player and can be downloaded separately as WAV or MP3, so you can take just the drums for sampling or just the vocal for a remix.
Can I listen to stems together before downloading?
Yes — after splitting you get a live mix where each stem has solo and mute buttons. Drop the vocal to rehearse karaoke, solo the bassline to learn it, or mute the drums to practice over the rest of the band.
Is the stem splitter free, and is my music uploaded?
It is free with no signup or watermark. The heuristic split runs inside your browser with the Web Audio API, so the song is not uploaded. Inputs are capped at 200 MB and practical track length depends on your device's available memory.