FULL STORY

Mirelo SFX 2.0: Consistent Sound Effects for AI Video

Mirelo released SFX 2.0 on October 14, a sound-effect model that generates frame-consistent audio for AI-generated video and removes leaked background music while preserving dialogue.

2026-10-14 ~ 2026-10-14 · 2 episodes · 13 posts

Episode 1 · Mirelo Launches SFX 2.0, Frame-Accurate Sound Design for AI Video (2026-10-14, 10 posts)

On October 14, Mirelo released SFX 2.0, a sound effect generation model that tackles the common pain point of AI-generated videos—great visuals that fall apart the moment sound is added—by automatically producing a complete sound design from video or text prompts.

Confirmed

  • The model works by watching the video: it automatically strips out background music the user didn't request in the prompt, preserves dialogue, and frame-by-frame generates sound effects, foley, and ambient audio synced to the on-screen action to build a complete soundscape
  • Its core selling point is cross-shot consistency: the same vehicle, character, or scene keeps a uniform timbre across different shots, and ambient sound doesn't break across cuts or arbitrarily long clips
  • Availability: launched on Mirelo's own API with Runway as a launch partner; also available on fal the same day, supporting both text-prompt and video input

Why it matters

  • Creator Jaynit Makwana demonstrated the typical problem with an AI-generated Clash of Clans battle video: every cannon blast sounds different, and ambient audio restarts after scene changes—ordinary viewers can't name the issue but clearly sense something is off. This represents the most pervasive audio-breakage problem in AI video assembly workflows
  • Sound consistency is seen as one of the key gaps between generative video being "watchable" and being "commercially viable"; if SFX 2.0 delivers reliably, it can plug directly into existing AI video production pipelines like Runway and fal, without manual re-dubbing or clip-by-clip audio fixes

Episode 2 · Mirelo SFX 2.0 Removes Unwanted AI Video Music While Keeping Dialogue (2026-10-14, 3 posts)

Mirelo's SFX 2.0 tackles the 'music leakage' problem in AI-generated video by selectively removing unwanted background music and re-adding ambient sound while preserving dialogue, winning praise from developers testing the workflow.