THE TRACK, UNLAYERED

Stem splitter: open a song into four parts

Pull vocals, drums, bass and other instruments into separate MP3 tracks. Preview each part in a simple mixer before you save it.

YOUR FILE, YOUR MIX

Separate a track

Drop in audio, choose what you want to hear, then preview and download your result.

3 free tracks / day

No file selected

Choose your output

Upload only. Use audio you have the right to process.

THE PROCESS

How it works

  1. 01

    Choose an MP3, WAV, FLAC, M4A or OGG file from your device. It must be no larger than 50 MB or longer than 10 minutes.

  2. 02

    Leave Four stems selected so the vocal, drum, bass and other tracks appear individually.

  3. 03

    Press Separate audio and keep the tab open until processing finishes.

  4. 04

    Use the in-page preview to mute or adjust parts, then download the individual MP3 stems you need.

A stem splitter estimates the separate musical parts inside a finished recording. Upload an audio file above with Four stems selected and download individual MP3 files for vocals, drums, bass and other instruments. The page also gives you a four-part preview so you can change the balance and decide which part you want. These are estimated stems, not the original studio tracks used to make the mix.

What is in each stem?

The Vocals stem aims to hold the lead voice and other sounds the model recognizes as voices. Backing vocals may land there too, but effects and reverb can spread into the music stems. The Drums stem targets the percussion and rhythmic hits. It can include some sharp transients from other instruments. The Bass stem focuses on low-pitched bass instruments; deep kick drums may overlap it. The Other stem holds the remaining material together, usually instruments such as guitars, piano, strings and synths.

This is a four-source split. It does not make separate guitar and piano files. A model called “six-source” exists in the broader Demucs project, but this site uses the four-source htdemucs model. The extra Instrumental and Drumless files available from other modes are mixes built from these four stems, not two more independently separated instruments. Naming that difference matters when you plan a session in your audio editor.

The vocal remover page combines Drums, Bass and Other into a backing track. The Acapella Extractor page puts Vocals first. This page leaves all four parts separate so you can choose your own combination.

Preview the balance, then download

After processing, press Play mix to hear the four stems together. Mute Vocals to hear the backing without it, or mute Drums to hear how the other parts carry the rhythm. Use a stem’s volume slider to bring it forward or push it back. The position control lets you jump to a busy part of the song, where the separation is easiest to judge. The browser preview changes what you hear here; it does not create a new combined download.

Download each MP3 stem individually. In a digital audio workstation, import them to separate tracks and align them at the same start time. The outputs came from one pass over one source file, so they share a timeline. If you need a custom combined file, adjust levels and export it from your editor. You can also use the preview as a quick check before doing that work.

For practice, muting Drums can leave room to play along as a drummer; muting Bass can let you hear the harmonic instruments while you learn a bass line. For remix preparation, isolating the voice can give you a starting point for a new arrangement. For listening study, compare the bass and drum stems to hear how the groove is built. None of these uses turns an unlicensed recording into material you can publish freely.

Why a stem can contain sounds from another part

A mastered song is one combined waveform. The separator cannot return to the microphone or instrument tracks that created it. Instead, it uses patterns learned from music to estimate which sounds belong to which source. If a synth bass and a kick drum overlap in pitch and time, some of each can appear in the other stem. A distorted guitar and a voice can also share harmonics. Reverb can linger in more than one track.

Use the clearest source file you have permission to use. A higher-quality source may preserve details the model needs, but it cannot undo a mix where two sounds are inseparable. Preview the loudest chorus, a quiet passage and any solo you care about. If a stem has artifacts, an editor may help with small repairs; the model cannot promise perfect isolation.

File, time and rights limits

The upload accepts MP3, WAV, FLAC, M4A and OGG audio files up to 50 MB and 10 minutes. Output is 320 kbps MP3 only. Three jobs per day are free for each visitor/IP; a shared network can share the limit. The page shows the real upload percentage, then a processing status without a guessed countdown. When processing capacity is unavailable, you get a clear message.

Inputs and stems live in private temporary storage. Access expires after one day, followed by storage deletion. There is no public stem library, account or link-based import. Only upload audio you own or are allowed to process. If you want to release a remix or sample, make sure you have rights to the composition and the recording; separating them does not change those rights. The privacy and terms pages explain the handling and use rules.

GOOD TO KNOW

Frequently asked questions

What are the four stems?

Vocals, drums, bass and other. Other is the remaining instruments together, such as guitars, keys and sounds the model did not assign to the first three.

Can it split piano and guitar separately?

No. This launch uses a four-source model. Piano and guitar normally land in the combined other stem, and some sound can leak between stems.

Can I hear the stems together before downloading?

Yes. The four-stem preview has a play button, a volume slider and a mute control for each part. It helps you decide which individual files to save.

What format are the downloads?

Each stem downloads as a 320 kbps MP3. Input files can be MP3, WAV, FLAC, M4A or OGG, but the site does not output WAV or FLAC.

Are the stems kept online?

They are available only temporarily through signed links, and access to input and output audio expires after one day, followed by storage deletion. Download what you need before then.