← All audio toolsSeparate stems

PITCHCHANGER / 18

Stem Splitter

Separate vocals, drums, bass and other instruments. Remix the balance

How to use this tool ↓
WebGPU required · up to 6 minutes · model ≈181 MB4 stems + mixes · WAV
Choose or drop an audio file
Choose audio to start separation

Requires WebGPU. Maximum duration: 6 minutes. The first use downloads an approximately 181 MB model. Audio is processed on your device; available memory and storage also limit file size.

Processing stays on this device. Engines and models download when needed. Privacy & usage

How to use this workflow

  1. Check the device requirements

    This tool requires a browser with usable WebGPU. Starting separation loads or checks an approximately 181 MB model plus runtime assets. There is no separate model-download button.

  2. Choose a recording and start

    Select a supported file no larger than 250 MB and no longer than six minutes, then select Split four stems. Available memory and browser storage can impose tighter limits. Try a short excerpt first.

  3. Listen and download

    Run separation, then use synchronized playback, mute and solo to inspect the stems. Download the individual WAV files or choose Drumless, Bassless, Instrumental or Acapella from the export choices.

What the four outputs contain

Vocals estimates singing and other vocal material. Drums estimates percussion, Bass estimates bass instruments, and Music contains the remaining instrumental material. These labels describe the model’s categories, not guarantees of clean isolation.

Swipe or scroll sideways to read all columns.

ExportIntended content
Drums / Bass / Music / VocalsOne estimated stem per WAV file
DrumlessBass + Music + Vocals
BasslessDrums + Music + Vocals
InstrumentalDrums + Bass + Music
AcapellaVocals

Listen for leakage and missing detail

Cymbals can spill into other stems; reverberant vocals may remain in the instrumental mix; an instrument sharing the voice’s frequencies can be partially removed. Solo a stem to inspect the problem, then listen to the intended mix as well.

The tool converts audio to a 44.1 kHz stereo model input. Model files may be cached in browser storage, and processing can use temporary local buffers. There is no CPU fallback for this four-stem tool. It is not a cloud job that continues after the page is closed.

Common questions

Can I extract guitar as a separate stem?

There is no dedicated guitar stem in this model. Guitar usually shares the Music output with other instruments.

Are the stem downloads MP3 files?

The splitter’s download choices are WAV. Use the audio converter afterward if you need MP3 and accept the additional lossy encoding step.

Explore all 18 audio tools