Stem Separation Models Compared
Cloud Standard separates a song into 2 or 4 tracks on our servers. Local 6 Tracks runs on your own computer and returns drums, bass, vocals, guitar, piano and the rest. Try either one at the top of this page.
Credits are charged only when a task succeeds. If it fails, they are returned automatically.
Loading your tasks…
Hear it before you upload
A 30-second song made for this site, already separated into tracks. No sign-in needed. Solo or mute any track and play them in sync.
- Vocals
- Drums
- Bass
- Other
Demo song and tracks: generated by SampLab Pro with AI (Mureka through WaveSpeed) and separated with the Local 6 tracks engine, for demonstration only. No licence is granted to reuse them. Rights and licensing
Where to go next
More ways to use stem separation
Stem Separation Models features
Tracks: 2 or 4 in the cloud, 6 on your computer
Cloud Standard gives vocals and accompaniment, or those plus bass and drums. Local 6 Tracks adds guitar and piano and keeps the leftover sound as an other track.
Speed: the cloud does not depend on your device
Cloud Standard runs on our servers, so a slow laptop makes no difference. Local 6 Tracks uses your own hardware: on our test laptop a 2 minute 43 second song took about a minute with GPU acceleration, and several minutes without it.
Devices: any browser for the cloud, a computer for local
Cloud Standard works wherever you can upload a file. Local 6 Tracks is made for computers: the first use downloads about 147 MB and the work needs a few gigabytes of memory.
Cost and privacy: credits or free, uploaded or on your device
Cloud Standard is charged per minute of audio and returns your credits if a task fails. Local 6 Tracks is free, needs a sign-in, and your audio never leaves the device.
Who uses Stem Separation Models
You want a quick instrumental or vocal
Pick Cloud Standard with 2 tracks. It returns the vocal and the accompaniment, which is all a karaoke track or an acapella needs.
You want to mute one instrument
Pick Local 6 Tracks. With the guitar, bass, drums and vocal on their own tracks, you can mute the part you want to play yourself.
You care where your audio goes
Local 6 Tracks keeps the song on your device from start to finish, which suits unreleased demos and recordings you cannot share.
You are on a phone or a slow machine
Pick Cloud Standard. The separation happens on our servers, so your device only uploads the song and plays the result.
Why choose Stem Separation Models
Credits only for finished results
A task is charged only when it succeeds. If it fails, your credits go back to your balance.
Both are in one workbench
Switch between Cloud Standard and Local 6 Tracks at the top of this page. Both let you preview, mute and download each track as WAV or MP3, or take all of them in one ZIP.
Honest about what each one does well
Local 6 Tracks is the only one with guitar and piano, but its piano track is the weakest of the six and can carry sound from other instruments. Cloud Standard stays with fewer tracks.
Your results stay private
What Stem Separation Models creates stays in your account until you download or save it.
Explore more tools
Every SampLab Pro tool is one click away.
AI creation
Edit audio
How to use Stem Separation Models
- 1
Choose a model
Open the workbench at the top and pick Cloud Standard or Local 6 Tracks. For the cloud, also choose 2 or 4 tracks.
- 2
Add your song and start
Upload a file or pick one from your library, then press the button. The cloud shows its price before you start; Local 6 Tracks shows progress and can be cancelled.
- 3
Preview and download
Play each track, mute the ones you do not want, then download them one by one or as a ZIP, in WAV or MP3 at 320 kbps.
Two ways to separate a song
SampLab Pro offers two stem separation models, and each one suits a different job. Cloud Standard sends your song to our servers and returns either 2 tracks (vocals and accompaniment) or 4 tracks (vocals, accompaniment, bass and drums). Local 6 Tracks never uploads your audio: it runs in your browser on your own computer and returns six tracks, drums, bass, vocals, guitar, piano and everything else.
This page puts them side by side, in plain terms, so you can pick the right one before you spend time or credits. The workbench at the top lets you try either model on your own song.
Which model for which job
Choose Cloud Standard when you want the instrumental or the vocal of a song and nothing else. Two tracks cover karaoke, a backing track and an acapella. Four tracks add the bass and the drums, which suits remixing or practising along with the rhythm section. A separation costs 9 credits per minute (minimum 36), charged only when it succeeds, and credits come back in full when a task fails.
Choose Local 6 Tracks when you want more detail than that: a guitar track to remove, a drum track to study, or a vocal you want to treat without uploading the song anywhere. It is free, and the limit is the length of your audio, up to 10 minutes, rather than credits.
What to expect from the six tracks
Local 6 Tracks returns drums, bass, other, vocals, guitar and piano. The vocal, drums and bass are the strongest. The guitar is usable. The piano is the weakest of the six and can contain bleed from other instruments, so listen to it before you rely on it. Everything the model cannot assign to a named instrument lands in the other track.
The first run downloads the model, about 147 MB, then keeps a copy in your browser so later runs start straight away. If your browser has no GPU acceleration, the page says so before it starts and the work can take several minutes. Separation is not perfect on any model, so faint traces of other sounds can remain.
Ready to try Stem Separation Models?
Choose Cloud Standard or Local 6 Tracks at the top of this page. Cloud separation costs 9 credits per minute (minimum 36), charged only when it succeeds; Local 6 Tracks is free.
Open Stem Separation ModelsStem Separation Models FAQ
What is the difference between Cloud Standard and Local 6 Tracks?
Cloud Standard separates on our servers and returns 2 or 4 tracks for a price per minute of audio. Local 6 Tracks separates in your browser, returns six tracks and is free. Local 6 Tracks never uploads your audio.
Which model should I use for a karaoke track?
Cloud Standard with 2 tracks. It returns the accompaniment without the vocal, which is the karaoke track, plus the vocal on its own.
Is Local 6 Tracks really free?
Yes. It uses no credits. You do need to be signed in, and the first use downloads about 147 MB to your browser.
Does Local 6 Tracks upload my song?
No. The song is decoded and separated on your device. Only the model files are downloaded; your audio is not sent anywhere.
How long does Local 6 Tracks take?
It depends on your device. On our test laptop a 2 minute 43 second song took about a minute with GPU acceleration and several minutes without it. The page tells you before it starts if your browser has no GPU acceleration.
Why is the piano track weaker than the others?
The piano is the hardest of the six sources for this model, and the model authors say it can leak other instruments. Treat the piano track as a rough guide rather than a clean stem.
How much does Cloud Standard cost?
Each task costs 9 credits per minute (minimum 36), based on the length of your song. Credits are charged only when a task succeeds; if it fails, they go back to your balance.