Diffcut
Clip RoomA long video or stream into 20–30 clips.EditingTrims, keyframes, transitions, export.CaptionsTranscribed on your Mac, word by word.Voiceover28 voices, with pauses where you write them.Studio SoundClean up a voice in one click.DubbingYour video in another language, in your voice.AI musicMusic and effects from a prompt.Background removalRemove, replace or blur, no green screen.CaptureA built-in browser for shots and recordings.Coding agentsClaude Code, Codex and more, as your editor.ShortsVertical shorts, scored with reasons.Demo videosA first cut from a browser recording.All features
In the app

A clean voice in one click.

Switch on Studio Sound and Diffcut removes background noise and room echo from a clip’s voice, on your Mac, then evens it out and sets its loudness.

The Studio Sound panel: switched on at 90% strength with EQ, Level and De-ess, a −16 LUFS target, a compare switch, and loudness from −21.8 LUFS before to −16.0 after.
Before Speech over café noise
Spectrogram of eight seconds of speech buried in café noise: the whole frequency range is filled.
After Studio Sound, default settings
Spectrogram of the same speech after Studio Sound: the noise is gone and the words stand out against quiet gaps.
On device
DeepFilterNet3, converted by Diffcut to run on your Mac
8.5 MB
for the model. No account, and nothing leaves the Mac
4 targets
−14, −16, −19 or −23 LUFS, or keep the clip’s level

Cleanup, then polish.

One switch on the clip’s Sound section. The rest is there when you want it.

  • StrengthHow much cleanup to apply, in percent.
  • EQCuts rumble and boxiness, and adds a little presence.
  • LevelEvens out loud and quiet words.
  • De-essSoftens harsh s sounds.
  • LoudnessLoud for social video, podcasts and YouTube, under music or broadcast.
  • CompareSwitch between the original and Studio Sound, or hold \ to hear the original as it plays.

What you hear is what you export.

Studio Sound is rendered once on your Mac and played through the same engine for preview and export. Change the strength and only the quick polish runs again.

Questions

Is my audio uploaded?

No. The cleanup and the polish run on your Mac, and the result is kept in your project.

Which model does it use?

DeepFilterNet3 (MIT), which Diffcut converted to Core ML from the official release and checked against the original’s own output. It is 8.5 MB and runs on the Mac’s CPU.

Is the original sound kept?

Yes. Studio Sound is an effect on the clip. Switch it off, or compare with the original at any time; your media isn’t changed.

Does it work on every clip?

On clips on the main track, including audio-only ones. A voiceover on a lane above or below the main track can’t take it yet.

Make the cut.

Diffcut is free, open source and early: the first Mac build isn’t out yet. Leave your email and we’ll write once, when it’s ready.