Come funziona Rimuovi voce Acapella Domande frequenti
Stem
Splitter
Split a finished song back into stems — music on one track, vocal on the other
MP3, WAV o FLAC · fino a 20 MB e 5 minuti · 25 crediti a traccia
How to split a song into stems
- Upload the mix. MP3, WAV or FLAC up to 20 MB and five minutes. A bounce, a stream rip or your own master — the splitter does not care where it came from, only how densely it was mixed.
- The model splits it. A separation network trained on real isolated stems reconstructs two tracks out of one. It is a reconstruction, not an un-mix: the original multitrack does not come back, it gets estimated.
- Balance the two stems. Both stems come up on their own fader with their own waveform, so you hear the split rather than reading about it. Vocal down is a karaoke instrumental, music down is an acapella, and anywhere in between is a rough edit you can audition before committing.
- Export the split. The export is the balance on screen, as MP3 or WAV — not two files you have to line up and recombine in something else.
What stem splitting is for
Domande frequenti
How many stems does it split into?
Two: music and vocal. That is a deliberate choice, not a limitation we forgot to lift. Four- and six-stem splitters give you drums, bass, other and vocal — and on most real jobs you immediately sum drums, bass and other back together to get the instrumental you actually wanted, having paid for three separations and inherited three sets of artefacts to get there. Splitting straight to two is cleaner, faster and cheaper for the thing people overwhelmingly do with it.
Is this the same as the original multitrack?
No, and no splitter is. The stems are estimated from the finished mix, which means every compressor, reverb and bus effect that was printed into the master is still printed into the stems. Real stems come off the session before the mix; these come out of it afterwards. For remixing and DJing the difference rarely matters. For a proper re-mix from scratch, ask the artist for the session.
Which songs split badly?
Three cases, consistently. Very dense, heavily limited masters, where there is little space between the parts for the model to work in. Heavy reverb and delay on the vocal, whose tails get read as music and left behind. And vocals processed until they stop sounding like a voice — vocoder, talkbox, hard formant shifting — which the model classifies as an instrument, correctly.
What does splitting a track cost?
Twenty-five credits a track, whichever page you start from — splitting, removing and extracting are the same separation with the faders in different places. The split is done once and both stems are yours.
Where do my stems end up?
In your own private library, alongside everything else you have made, ready to re-download for as long as your account is open. Uploads are sent securely for processing and are never published.
Una sola separazione, sei porte d’ingresso
These are all the same AI separation with the faders in a different place. Whichever one you start from, the split runs once and you get both tracks.