Diffcut
Get early access
EditingTrims, keyframes, transitions, export.CaptionsTranscribed on your Mac, word by word.Voiceover28 voices, with pauses where you write them.AI musicMusic and effects from a prompt.CaptureA built-in browser for shots and recordings.Ask ClaudeYour own Claude proposes edits you review.ShortsVertical clips from a long video.Demo videosA first cut from a browser recording.All features
In the app

Captions, word by word.

Diffcut transcribes on your Mac’s Neural Engine with a time for every word, then lays captions out to reading speed and clear of shot changes. Nothing is uploaded.

Get early accessSee the languages
On device
Parakeet TDT 0.6B v3 with Silero VAD, through Core ML
37 ms
to lay out 20,000 words around 4,000 shot changes, down from 3.5 s
5 styles
Stream, Broadcast, YouTube, Vertical and Word pop

Five styles, five reading speeds.

Each style sets line length, lines on screen and how fast people are asked to read. Pick one and the whole track follows it.

  • StreamTwo lines of 42 characters, for streaming and long programmes.
  • BroadcastTwo lines of 37 characters, slower, as European broadcast does.
  • YouTubeTwo lines of 42 characters, for long-form online video.
  • VerticalShort lines of 24 characters for 9:16 video.
  • Word popOne to three words at a time, each word lighting up as it’s said.
  • Your edits stayRe-run after a cut and only the stretch around it changes.

Laid out like a subtitler would.

Captions break at pauses and speaker changes, stay up long enough to read, and keep their edges clear of shot changes, to the frame.

  • Out as files, or in the pictureExport SRT, VTT or ASS, or burn captions into the video with word highlighting and emphasis pops.
  • Same words, same captionsLayout is deterministic: the same words and settings always give the same track.

25 languages, on your Mac.

English and 24 more European languages, or let Diffcut detect the language for each stretch of speech.

  • Bulgarian
  • Croatian
  • Czech
  • Danish
  • Dutch
  • English
  • Estonian
  • Finnish
  • French
  • German
  • Greek
  • Hungarian
  • Italian
  • Latvian
  • Lithuanian
  • Maltese
  • Polish
  • Portuguese
  • Romanian
  • Russian
  • Slovak
  • Slovenian
  • Spanish
  • Swedish
  • Ukrainian

Planned Chinese, Japanese, Korean and Hindi, with a second speech model. They aren’t in the app yet.

Questions

Which languages can Diffcut transcribe?

25 today: Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, Ukrainian. It can also detect the language for you. Chinese, Japanese, Korean and Hindi are planned with a second speech model.

Is my audio uploaded?

No. Transcription runs on your Mac’s Neural Engine through Core ML. The audio never leaves the machine.

Can I export SRT or VTT files?

Yes: SRT, VTT and ASS. Or burn the captions into the picture, with word highlighting, when you export the video.

Can I translate captions?

Through your own Claude: Diffcut’s MCP server has tools that prepare a caption track for translation and apply the result. There is no built-in translation model.

More of Diffcut

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.