Stem Separation — 4 Stems (Vocals, Drums, Bass, Other)
Stem separation takes a finished song and splits it into the parts it was mixed from: vocals, drums, bass and everything else. It is the difference between having a song and having the ingredients of a song.
Producers use stems to remix, sample, practise and rebuild arrangements. Upload a track below and get all four stems back in minutes. Runs are paid — the one-time $2.90 Trial Pack includes 15 minutes and WAV output.
By StemSplit Audio TeamLast updated: Editorial standards
Trial Pack: 15 minutes for $2.90, one-time payment. Runs start after checkout.
Paid runs only · $2.90 Trial Pack includes 15 minutes · files deleted automatically after processing
Four stems, one upload
Every job returns the full set of separations, not just a karaoke-style instrumental. That means one upload gives you an acapella, a backing track, a drum loop and a bass line you can drop straight into your DAW.
What comes back in your download
Lead and backing vocals, harmonies and ad-libs, isolated so you can re-sing them, sample them or build an acapella.
Kick, snare, hats and percussion as one clean stem for remixes, drum practice and loop-based production.
Bass guitar and sub-bass separated from the mix, ready for sampling, transcription or replacing the low end.
Guitars, keys, strings, synths and background textures — every element that is not vocals, drums or bass.
The instrumental is simply your original track with the vocal stem removed — it is created in the same pass, so you get vocals, instrumental and the individual stems from a single upload.
What stem separation actually gives you
A mixed song is a stereo file — two channels of audio with every instrument and voice baked together. Stem separation estimates how much of that signal belongs to each musical source and rebuilds those sources as separate files. The four standard stems are vocals, drums, bass and other, where 'other' covers guitars, keys, strings, synths and any remaining texture.
The stems are time-aligned with each other, so dropping all four into a DAW at bar one reconstructs the original song almost exactly. That is what makes them useful: you can mute the vocal, replace the drums, loop the bass or process one element without touching the rest.
- Vocals — lead, harmonies and backing vocals as one stem.
- Drums — kick, snare, hats and percussion.
- Bass — bass guitar and sub-bass.
- Other — guitars, keys, strings, synths and textures.
- Instrumental — the original mix minus the vocal, included in the same job.
How it works
First, get a paid run: the $2.90 Trial Pack is a one-time 15-minute allowance, and subscriptions add monthly minutes. Then upload an MP3, WAV, M4A or FLAC file. The AI model analyses the mix and estimates each source across the frequency spectrum, then renders the stems. A three-minute song typically takes around a minute, and the result panel lists every file with a download link.
There is nothing to configure. The model returns the same four stems for every genre, at the original tempo and key. If you need more granular separation — for example splitting the 'other' bucket further — the Pro plan adds a 6-stem model in addition to the standard output.
What producers do with stems
Remixers rebuild the arrangement around an existing vocal or drop a familiar drum groove under a new track. Sample-based producers isolate a single bar of bass or a drum break and chop it. Educators and students mute the part they are learning. Video editors remove an instrument that clashes with a voiceover without losing the rest of the music.
The common thread is control. Once a song is stems, every production decision becomes reversible: you can rebalance, replace, edit or process any element without re-recording anything. That is why stem separation has become a standard first step in remix and sampling workflows.
- Remix and mashup production.
- Sampling and chopping individual parts.
- Instrumental and acapella practice tracks.
- Music for video where one element conflicts with dialogue.
- Educational transcription and part learning.
Stem separation vs. vocal removal
A vocal remover is a single-purpose view of the same technology: it focuses on the vocal-versus-instrumental result. Stem separation exposes all the parts. In practice one job gives you both, because the instrumental is derived from the stems and the acapella is simply the vocals stem.
If you only need a backing track, use the vocal remover pages and ignore the extra files. If you are producing, mixing or sampling, go straight to stems — the individual parts are what you will actually work with.
Getting production-ready stems
Start with the highest-quality source you have. Lossless WAV or FLAC separates better than a low-bitrate stream, because compression smears sources into each other before the model ever sees the file. Every paid run delivers WAV as well as MP3, so you can mix with the lossless files.
Expect the practical limits of the model: some vocal bleed in the instrumental, occasional cymbal artifacts in dense mixes, and imperfect separation on live or mono recordings. For critical production work, use the Pro high-quality model and A/B the stems against the original mix before committing.
- Upload lossless or 320kbps sources.
- Download WAV for mixing and mastering.
- Check phase and timing against the original.
- Use the high-quality model for dense arrangements.
- Keep the original file as your reference master.
Formats, limits and plans
The one-time $2.90 Trial Pack is the entry point: 15 minutes, all four stems, MP3 320 and WAV, no subscription. Fifteen minutes is about five three-minute songs, and the minutes never expire. Starter ($5/month) adds 90 minutes per month and batch upload; Pro ($9/month) adds 200 minutes, the 6-stem model, high-quality mode, priority processing and a commercial-use license for the service. Annual is $39 — five months of Starter for a year.
Every paid plan uses the same separation engine — you are paying for volume and output options, not for a different model on a free-of-charge allowance. There is no complimentary tier.
Sample outputs are illustrative; real A/B samples will be added after launch. No customer counts, reviews or benchmark numbers are published on this page because we have not measured them yet. Every upload is processed only for your job and removed from working storage automatically within 24–48 hours — read the privacy policy.
System built-in method: Windows, macOS, phone and DAW
Built-in tools cannot produce stems. Windows, macOS, iOS, Android and every stock DAW offer at most a center-channel cancel — one mono instrumental, not four separated files. Here is the honest platform-by-platform picture.
Windows 11 / 10
Windows has no built-in vocal remover. Voice Recorder and Clipchamp can record, trim or change volume, but they cannot separate or cancel a vocal.
- Install Audacity from the official site (audacityteam.org). It is free and open source, but it is not bundled with Windows.
- Open your song as a stereo track (File → Open). Mono files have no center channel to cancel.
- Select the whole track (Ctrl+A), then open Effect → Vocal Reduction and Isolation.
- Choose Action: Remove Vocals, keep the default settings and apply.
- Check the result on headphones: centered bass and kick are usually removed together with the vocal.
- Export with File → Export Audio and choose WAV or MP3.
Good for a rehearsal backing track. It cannot keep the low end and it cannot produce an acapella or stems.
macOS
macOS ships GarageBand, but GarageBand has no phase-invert plug-in — so the bundled app alone cannot cancel a centered vocal. QuickTime Player can trim and export audio, not separate it.
- Open the song in GarageBand (bundled) if you only need to trim, split or adjust level — that works.
- GarageBand's EQ can notch 1–4kHz to reduce vocal presence, but it also thins guitars, cymbals and snare.
- For actual center cancellation, use Audacity (free, cross-platform) and follow the Windows steps above.
- Logic Pro (paid, not bundled) includes the Gain plug-in with Phase Invert: split stereo, invert one side, then sum to mono.
- Listen in mono afterwards — phase tricks can sound different when a phone speaker folds the mix down.
Honest answer: macOS has no bundled vocal remover. GarageBand EQ is a partial workaround, not a separator.
iPhone & iPad
iOS and iPadOS have no built-in vocal remover for your own files. GarageBand for iOS can record and arrange, but it has no phase-invert control. Apple Music's Sing feature can lower vocals on supported catalog songs, but it streams only and cannot export an instrumental.
- Apple Music Sing: play a supported song, open the lyrics view and use the vocal slider to lower the vocal (Apple Music subscription required).
- There is no export: Sing is a playback feature, not a file tool, so it cannot give you a WAV or MP3 backing track.
- GarageBand for iOS can trim and arrange audio you import, but it cannot cancel or remove a vocal.
- iMovie can lower a clip's volume, not isolate a source.
- For real separation, open this site in Safari on the phone, or move the file to a Mac/PC and use a desktop method.
Phones are for listening and file handling. No iOS feature produces stems or a downloadable instrumental.
Android
Android has no system vocal remover. Files by Google, Google Photos and Google Drive can move or trim media, but none of them separate audio sources.
- Google Files and Drive can move the audio between apps, but cannot edit or separate it.
- Third-party Android 'karaoke' or 'vocal remover' apps usually run the same center-channel math and damage the mix.
- If an app asks you to upload the track to its server, read its privacy policy first — many keep files far longer than they say.
- For AI separation, open this site in Chrome on Android; the upload and download flow works in the browser.
- For manual phase cancellation, move the file to a desktop with Audacity.
Android's built-in apps do not separate stems. Treat app-store vocal removers as phase filters unless they clearly run an AI model.
DAW (Audacity, GarageBand, Logic Pro, REAPER)
Every DAW can perform center-channel cancellation manually if it supports channel splitting and phase inversion. This is the real built-in technique — no plug-in purchase required.
- Import the song as a stereo track and split it into two mono channels.
- Invert the phase (polarity) of one channel only.
- Sum both channels down to mono. Anything centered — usually the lead vocal — cancels out.
- In REAPER, use the item phase (polarity) button before summing; in Logic Pro, use the Gain plug-in's Phase Invert.
- Audacity's Effect → Vocal Reduction and Isolation automates the same math.
- Check the result on headphones: if the bass and kick vanished, the method worked as designed — that is its limit.
Instant and offline, but the output is a mono backing track, not stems. There is no acapella and no preserved low end.
Built-in method vs. AI separation
If your goal is vocals, drums, bass and other as separate, time-aligned files, no built-in OS or DAW feature can do it. AI stem separation is what returns those four stems; the $2.90 Trial Pack includes 15 minutes with WAV and MP3 downloads.
Documentation we check for these steps: Audacity manual · Apple GarageBand · Apple Logic Pro · REAPER · Demucs (Meta AI) on GitHub · Replicate: cjwbw/demucs.
Frequently asked questions
Is there a free stem separation option?
No. There is no free-of-charge allowance and no complimentary preview: every separation costs GPU time, so every run is paid. The cheapest entry point is the one-time $2.90 Trial Pack — 15 minutes, about five songs, with all four stems in WAV and MP3.
How many stems do I get?
Four by default: vocals, drums, bass and other. The instrumental and the acapella are included in the same job. Pro adds a 6-stem model that splits the 'other' bucket further.
Are the stems time-aligned?
Yes. All stems are rendered from the same analysis and line up at the same start point, so importing them into a DAW at the same timestamp reconstructs the original song.
Can I get 6 stems instead of 4?
The Pro plan includes the 6-stem model in addition to the standard 4-stem output. It is useful when you want more detail in the guitar, piano or 'other' content.
What format are the stems?
Every paid run delivers MP3 320kbps and WAV — the Trial Pack included. Use WAV if you plan to mix, master or process the stems further.
Can I use separated stems commercially?
We license the separation service to you, but the rights to the source song belong to its owners. If you own or have licensed the track, you can use the stems commercially. If not, publishing or monetising them may infringe copyright.
Separate your first song into stems
Buy the $2.90 Trial Pack, upload a track and download all four stems in about a minute. 15 minutes, one-time payment, WAV included.
Read how the stem separation works, how your audio is handled, or contact the team.