TUNENOODLE / Stem splitter / Vocals
Ensemble vocal separation: compare another vocal result
Use the MVSEP ensemble alternative to separate vocals and instrumental, then compare difficult passages by ear.
Ensemble vocal separation: compare another vocal result
Some songs need a second separation attempt because the first result loses a quiet syllable or leaves a bright instrument in the vocal. Ensemble vocals gives you an alternative algorithm for that comparison, with two outputs: vocals and instrumental. In source separation, an ensemble combines contributions from multiple models rather than treating one model's estimate as the only answer. That can change the balance between preserving a voice and suppressing the surrounding music, but it does not guarantee a better result on your particular song. The most useful comparison is a short, difficult passage: a breathy verse, layered chorus, consonants beside hi-hats, or a held note over guitar. Keep playback levels similar and listen for what was retained as well as what was removed. For a remix vocal, an intact phrase may matter more than a perfectly silent gap; for a backing track, a stable instrumental texture may matter more than eliminating every trace of reverb. This mode does not create separate singers or lead and backing tracks. It provides another two-track estimate to evaluate against the standard vocal splitter before you commit to an edit.
How to separate your track
- 01
Choose ensemble vocals and load the same source used for your comparison.
- 02
Audition vocals and instrumental, focusing on the difficult phrase.
- 03
Keep the version whose retained detail and artifacts suit your actual use.
What this mode gives you
| Parameter / option | What it changes |
|---|---|
| Start with the right source | Use the same source and timestamps as your first vocal separation. Avoid comparing different masters or louder playback, which can obscure the algorithm difference. |
| Good for | Recover a more useful version of a quiet vocal phrase Compare the instrumental texture around a dense chorus Choose between vocal fullness and distracting bleed for a particular edit |
| Listen for | Compare whether quiet consonants and sustained notes remain intact. Listen for changing guitar or cymbal leakage between words. Check instrumental continuity at the same vocal entry in both versions. |
A few useful answers
No. It produces a different estimate. Compare the passage you need at similar playback levels.
Your next step
All modesUses credits