Skip to content
Bernhard GötzendorferAI automation that runs on its own · Vienna

AgenticCutter

You read the rough cut as text and approve it.

Mac app · early version · agentic-cutter.com (opens another website)

AgenticCutter, window recreated: list “Pauses” with timestamps, a reason such as Filler, Silence or Noise, and a switch showing “Cut” or “kept” on each row
Window recreated

What you do with it

You have recorded a short clip with two ums, six pauses and one slip: “Let me say that again.” You drop it into the cutter, and it transcribes the clip on your Mac. The agent marks the spots; you go through the list and keep one pause because it lands well. Then you approve, and only then does it render.

Who it's for
For anyone who edits their own videos and would rather read text than click through waveforms.
Where it stands
An early version for Macs with Apple silicon. Transcription and cutting run on your Mac, and you decide what, if anything, goes to the cloud.

Step by step

  1. Import your video

    You drag your video into the app. To try it out first, you can open the sample project instead.

  2. Transcribe locally

    The cutter writes down what you say, word by word, on your Mac. The text then shows you where the cuts go.

  3. Go through the suggestions

    The agent suggests cuts, at filler words and pauses for instance. You switch each one to cut or keep.

  4. Approve, then render

    Nothing renders until you approve it. Without your approval there is no export.

How it works

The language model
reads the transcript and suggests cuts.
The code
transcribes with mlx-whisper and cuts with ffmpeg. Your video is not uploaded for editing.

Why it exists

Turning an hour of footage into a video cost me two to three hours at the desk. A machine finds the pauses and filler words, and I decide what stays in.

Further reading

Other things I have built