Choose your song
Upload MP3, WAV, FLAC, or M4A.
Acapella maker
Upload a song and hear the opening excerpt of the predicted vocal stem. Separation quality varies, so use the free preview as an initial check before unlocking the complete track.
Separation job
Outputs vocals, drums, bass, and other.
Uploads go to R2, processing runs through Replicate Demucs, and downloads unlock after sign-in.
Upload MP3, WAV, FLAC, or M4A.
Secure cloud processing separates instruments from the vocal; the exact time depends on the track and queue.
Free 30-second preview; full acapella in WAV, FLAC, or MP3 with credits.
These licensed source recordings were AI-separated by StemSlice and excerpted to 30 seconds. User uploads are private and never published.
Pop / 0:30
Switch tracks to compare the same point in the 30-second excerpt.
"Bummin on Tremelo" by Kevin MacLeod (incompetech.com), licensed under CC BY 4.0.
Creative Commons Attribution 4.0 license · AI-separated by StemSlice and excerpted to a 30-second preview.
Contemporary / 0:30
Switch tracks to compare the same point in the 30-second excerpt.
"Ether Vox" by Kevin MacLeod (incompetech.com), licensed under CC BY 4.0.
Creative Commons Attribution 4.0 license · AI-separated by StemSlice and excerpted to a 30-second preview.
Jazz / 0:30
Switch tracks to compare the same point in the 30-second excerpt.
"Secret of Tiki Island" by Kevin MacLeod (incompetech.com), licensed under CC BY 4.0.
Creative Commons Attribution 4.0 license · AI-separated by StemSlice and excerpted to a 30-second preview.
Rock / 0:30
Switch tracks to compare the same point in the 30-second excerpt.
"Exhilarate" by Kevin MacLeod (incompetech.com), licensed under CC BY 4.0.
Creative Commons Attribution 4.0 license · AI-separated by StemSlice and excerpted to a 30-second preview.
Electronica / 0:30
Switch tracks to compare the same point in the 30-second excerpt.
"Townie Loop" by Kevin MacLeod (incompetech.com), licensed under CC BY 4.0.
Creative Commons Attribution 4.0 license · AI-separated by StemSlice and excerpted to a 30-second preview.
Jazz / 0:30
Switch tracks to compare the same point in the 30-second excerpt.
"Faster Does It" by Kevin MacLeod (incompetech.com), licensed under CC BY 4.0.
Creative Commons Attribution 4.0 license · AI-separated by StemSlice and excerpted to a 30-second preview.
Choose the passage first, then evaluate timing, artifacts, and transformation risk.
Pop, hip-hop, rock, EDM — if there's a lead vocal, the AI can find it.
Pair your acapella with the isolated drums and bass from the same job.
One upload, no audio engineering required.
Each job expires 24 hours after creation and is then removed by automated cleanup; final physical removal timing can vary.
Keep the source, untouched extraction, intended use, and permission status together.
Find the hook, verse, ad-lib, or harmony stack that will actually be used, then judge that passage in the preview. There is little value in over-processing an entire vocal when the production needs four bars. Keeping edits purpose-specific also makes timing, fades, and permission records easier to manage.
First align the original vocal and confirm its tempo. Then apply key or tempo changes in the DAW, using formant preservation when appropriate. Large transformations magnify separation artifacts, so compare a small test render with the untouched extracted stem before committing to a full arrangement.
The tool can produce a workable vocal file but cannot grant a license. A recognizable lyric, melody, or performance may require permission even when it is chopped, pitched, or layered. For released music, keep a record of the source, rights holder, license, and approved use.
Record the song, version, timestamp, intended edit, and permission status before pitching or chopping the vocal. Large time and pitch changes can hide some artifacts while exaggerating others. Keeping the untouched extracted stem beside the transformed version makes both quality review and rights clearance easier.
Take the intended four or eight bars through the actual pitch, time, EQ, and compression chain. Separation artifacts can become more obvious after formant shifting or heavy limiting, even when the untreated preview sounds acceptable. A short test saves conversion time and shows whether another source section would survive the production better.
1 credit unlocks one full 4-stem download. Previewing, listening, and failed processing jobs do not consume credits.
$4.9
5 credits / one-time
Buy Credits$9.9
30 credits / one-time
Buy Credits$19.9
60 credits / one-time
Buy CreditsAnswers for free limits, formats, privacy, copyright, and download quality.
Upload the song here. The AI separates the vocal from the instruments and gives you the acapella as a downloadable file.
Making and previewing acapellas is free (3 songs a day). Downloading the full-quality version costs 1 credit, from $4.9 for 5.
Remixes, mashups, DJ tools, vocal practice, and covers are the common ones. Commercial releases need permission from the rights holder.
It works best on studio recordings with a clear lead vocal. Very lo-fi or live recordings can produce artifacts.
Two names for the same job: getting the vocal out of a finished song.
Up to 6 minutes free, 20 minutes for paid users.
Files are encrypted, set to expire 24 hours after job creation, and then removed by automated cleanup; final physical removal timing can vary.
Extracted stems remain the property of the original rights holders. Commercial use requires their permission.