Mirelo SFX 2.0 rebuilds the sound of a video and keeps the dialogue
Mirelo SFX 2.0 is a video-to-sound model that removes unwanted music from a clip, keeps the dialogue, and generates effects, foley and ambience synced to the action. It handles edits of up to 10 minutes in one pass and is now also available on Runway.
What it does
You upload a finished video. The model detects and keeps the dialogue, removes the other audio, such as music a video model added without being asked, and generates sound effects, foley and ambience that follow the action, fit the setting and duck around the speech.
Because it works on the assembled edit, the sound stays coherent across cuts, so several shots of one place sound like one sequence instead of separate generations.
Results and limits
In Mirelo's own blind listening tests (25 raters judging 30 videos per comparison), SFX 2.0 was preferred over Sony's Woosh model in 81% of cases and rated as good as or better than Seedance 2.5's native audio in 72% of cases. These are vendor-run tests.
Worth knowing
It's available in the Mirelo Playground, Editor, API and MCP server, and on Runway and fal. Mirelo also lists plugins for Adobe Premiere Pro, DaVinci Resolve and Roblox, with some marked as coming soon. Mirelo Studio and the API take videos up to 10 minutes, with longer videos on the Enterprise tier.