
Claude Cut is built for single-speaker footage: a script read to the camera, YouTube monologues, founder explainers, course modules, interview answers. It is not a multicam or dialogue-scene tool.
Quick Start
1
Choose the range
Open Claude Cut from the Tools Menu and pick Whole timeline, In / Out or Selected clips (Premiere Pro). Below, tick the media whose speech is transcribed, and choose the tracks to cut if your timeline has several.
2
Add a brief or your script (optional)
In Settings, write a short brief or paste the script you read to the camera. Both help the AI keep the right takes.
3
Click Analyze
The green Analyze button runs everything on its own: silence detection, transcription, AI analysis. You watch the AI judge each sentence live.
4
Tune and review
Set a target duration if you want a shorter video, adjust the sensitivities in Advanced Settings, and check the result on the timeline preview and in the transcript.
5
Apply Cut
Click Apply Cut. The cut parts are removed and the gaps closed, in one click.
What Claude Cut removes
Off-camera talk
Mic checks, countdowns, “is it recording?”, words to the crew, “sorry, let me redo that”, chit-chat once the content is over. Greetings and calls to subscribe addressed to your audience are kept.
Retakes
False starts, lines read with a slip and read again, a first wording you then said differently. The last take is always kept: Claude Cut only cuts a sentence when a later one redoes it, so a line can never disappear entirely.
Fillers
“Um”, “uh”, “euh”, “äh”, “えーと”… in the language of your video. Words that also have a real meaning in that language (like “um” in Portuguese or “like” in English) are never cut.
Silences
The dead air between your sentences, from a minimum duration you choose.
What it never does
- Rewrite or reorder anything. Your words stay in your order. The only lever is which parts survive.
- Cut a sentence that adds information, even when it looks like the previous one.
- Cut a deliberate repetition. Rhetorical triples and emphasis (“very, very fast”) stay.
Step 0: Range & media

Range
- Whole timeline analyses the sequence from start to end.
- In / Out only the range between your In and Out points.
- Selected clips (Premiere Pro) only the time covered by the clips selected in the timeline, on their tracks only: the tracks to cut are set to the tracks of your selection.
Media used for transcription
Below the range, tick the media whose speech is transcribed. Each media is transcribed once, in full, kept on this computer, then its words are placed back on the timeline wherever its clips are. A media already transcribed by any PremiereGPT tool (Smart Subtitles, Smart Profanity, Auto Zoom…) is reused: no new transcription and no transcription fee, only the AI analysis runs. By default, the media already transcribed are ticked and the new ones are not. See Subtitles Pro for the details of the list.Tracks to cut
When your timeline has at least three tracks with clips, a row of short buttons (V1, V2, A1, A2…) lets you choose the tracks Claude Cut cuts. All of them are on by default; the full name of each track is in its tooltip.- Only the ticked tracks are cut and shortened. The other tracks (music, ambience, another speaker) keep their timing.
- Only the speech on the ticked audio tracks is read: untick a music or second-speaker track and its words are ignored by the analysis.
Pick at least one track: with none ticked, Claude Cut can’t run.
Step 1: Settings
Brief
A few words of intent, for example “keep the jokes” or “I restart my sentences a lot”. Keep it to two or three sentences: long rule lists don’t produce a better cut.Script
If you read a script to the camera, paste it here (the paste button sits in the corner of the box). The AI uses it as the reference for what you meant to say, which helps it keep the right take and spot what is off script. You can read it again during the review, in the Script view.
Transcription precision
The small chevron to the left of Analyze opens the engine menu:PremiereGPT (standard)
Fast, reliable transcription. The right choice in the vast majority of cases.
+ Advanced word precision
Re-syncs every word to the audio for sharper cut points. Slightly slower and costs extra credits (shown on the button).
The analysis
The Analyze button chains three steps, shown live:
While the AI works, the window shows each sentence in a triangle: one corner per question, a point that moves with the answers, and a green wave when the sentence is kept. At the bottom, the time removed and the number of sentences kept, retaken and off camera grow live.

A run from Analyze is billed as one combined charge, and only if it completes and finds something to cut. When the words come from the media list, that charge no longer includes the transcription: new media are billed per minute once, as shown under the list, and media already transcribed cost nothing more. Clicking Re-analyze only bills the analysis.
Step 2: Review
Every setting here recomputes the cut instantly, without a new analysis.Target duration
Drag the slider to the length you want. Claude Cut first removes the cuts above, then, if the video is still too long, the least essential sentences first, until it fits. At the far right, No limit keeps everything the AI didn’t flag.Advanced Settings
The count next to each setting tells you how many sentences or words it affects.

Timeline preview
Every sentence appears on a timeline, colored by what happens to it: kept, off camera, retake, cut for the target duration, or cut by hand. The silences to cut sit on a thin line below. Click a sentence to keep it or cut it: the playhead jumps there and the transcript scrolls to it.
Transcript
Three views:- Edit: every sentence with its KEEP / CUT button and, when it is cut, the reason.
- Kept text: the same layout, without what is cut. This is what your video will say.
- Script: the script you pasted in Settings, if any.

At the bottom, Final duration shows the length before and after the cut, pauses included, and how much is removed. It updates as you move the sliders.

