Your video, in your voice, in another language.
Diffcut speaks every line again in a new language, in the speaker’s own voice cloned on your Mac with their consent, timed to the picture and over the original music.


- 10 languages
- from Spanish and Japanese to Chinese and Russian
- ~50 s
- to dub 70 seconds of video on an M5, once it is transcribed
- 1.2× at most
- a line is sped up to fit, pitch kept; longer ones are rewritten
Only with consent.
The Dub sheet plays the exact stretch the voice is cloned from. You confirm it’s your voice, or that you have the speaker’s permission. Only that choice makes a voice; an agent can ask for one, never confirm it.
- From your footage onlyA voice is cloned only from the project’s own footage.
- Watermarked and labelledEvery dubbed line carries a verified watermark and is marked AI-generated.
- Withdraw at any timeWithdrawing a voice deletes it; no new line can be spoken in it.
Timed to the picture.
Each line starts where the original did and fits the room it had, its own length plus the pause after it. Lines that run long are sped up a little, then rewritten shorter by your Claude.

Rewrite a line and speak it again with ⌘↩, have it shortened, or move it by a tenth of a second. Every change is one undo.
It keeps the music.
Diffcut separates the original’s music and effects from the voice on your Mac, so the dub plays over them. The dub opens as a version of its own, with subtitles, and your edit stays as it is.
- Keep musicThe original voice out, its music and effects under the dub.
- Lower voiceThe original, 18 dB down under each line.
- RemoveOnly the dub.
- Subtitles includedOf what the dub says, as SRT, WebVTT or ASS.
Questions
Which languages can it dub into?
Spanish, French, German, Italian, Portuguese, English, Japanese, Korean, Chinese and Russian.
Who translates?
Your own Claude Code, through Diffcut’s tools, with a length budget for every line. An agent can also send its own translations.
Does the mouth match the new words?
Not yet. There is no lip-sync, so the mouth still moves to the original words.
What if more than one person speaks?
A dub uses one cloned voice for now, so every line is spoken in it. Or choose a preset voice.
What leaves my Mac?
The lines to translate go to your own Claude under your account. The voice is cloned, and every line spoken, on your Mac.
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.