Tools · Breakthrough Workflows
Curious Refuge found a way to remove the score without losing the footsteps
A tutorial on using SAM Audio to remove music from AI videos and regular footage, allowing for clean sound effects and custom music addition
Reporting updated Nov 11, 2026
THE TAKEAWAY
When a generated clip includes music the director did not request, the useful trick is separating it without destroying the rest of the sound.
WHAT HAPPENED
Curious Refuge demonstrates sending a generated clip into Meta’s SAM Audio playground, requesting removal of the music and using the residual output for dialogue, footsteps and ambience. The tutorial notes that prompt descriptions and source material can affect artifacts, and the edited audio may still need repair and synchronization. This is a documented workflow, not a claim that the tool can recover every sound flawlessly.
WHY IT MATTERS
Video models are becoming better at making audio along with pictures, but that also means unwanted sound can be baked into otherwise useful footage. The tutorial gives creators an immediately testable workaround.
Reader card
Choose the unwanted source
use SAM Audio separation
inspect the residual
repair and sync in an editor
VEYR ANGLE
Put the concrete before/after result first and identify SAM Audio as the separator, not the model that generated the original shot.
