TUNENOODLE / Stem splitter / FX & Dialogue

BandIt Plus soundtrack separator

Separate Speech, Music and Sound effects with BandIt Plus, then compare the layers around the words and scene events that matter.

BandIt Plus soundtrack separator

BandIt Plus is a cinematic separation option that returns Speech, Music and Sound effects. Its research lineage is the 2023 BandIt system: a model that analyzes frequency bands with shared information before reconstructing the different soundtrack roles. MVSEP identifies its Plus model with the original authors’ work and the Divide and Remaster dataset. That explains its purpose; it does not establish a universal quality advantage for every recording. Choose it when you want another interpretation of a difficult film, interview or documentary mix. A meaningful comparison uses the same source excerpt and checks the same word, cue and effect, rather than switching files or judging only the amount of background left in Speech. Keep the original timeline so differences can be heard at the same instant. Pay attention to the balance between clean isolation and complete content. Soft consonants, short musical notes and effects under speech can be lost even when the output sounds quieter. Listen to all three stems before assembling an edit, because useful material may have landed in a neighboring track. Plus names the model option; it does not mean extra output tracks, speaker separation or a guaranteed upgrade over the other available modes.

How to separate your track

  1. 01

    Select BandIt Plus and submit the original soundtrack excerpt.

  2. 02

    Compare all three outputs around an easy line and a heavily overlapped line.

  3. 03

    Use the result whose retained content best serves the edit, judging the same passage at comparable listening levels.

What this mode gives you
What this mode gives you
Parameter / optionWhat it changes
Start with the right sourceReuse the identical source excerpt when comparing models. Include one easy spoken line and one line under a cue or effect, so a clean easy section does not hide failure on the actual edit point.
Good forCompare a masked line of dialogue against another three-stem result. Recover a more useful music or effect layer when the first separation weakens it. Study how a scene element is assigned across Speech, Music and Sound effects.
Listen forDoes clearer speech come at the cost of consonants or breath continuity? Are musical notes intact during overlapping dialogue? Does the important scene effect remain in the effects track for its full duration?

A few useful answers

It is the model name, not an extra-stem package. The outputs are Speech, Music and Sound effects.

Your next step

All modes

Send feedback

Tell us what went wrong or what would help. Every message is read.

Type

Sending from /features/stem-splitter/bandit-plus