Video editors, worry: Claude Code cut a whole vlog from an hour of raw footage
Pieter Levels dropped 14 clips into a folder and asked Claude Code for a vlog. The AI transcribed the speech, scanned the frames, cut the video and drew captions on its own.

Pieter Levels, the indie hacker behind Nomad List and Photo AI, posted his latest vlog and only afterwards admitted he hadn't edited it. He dumped the clips from his DJI camera into one folder, opened Claude Code with Opus 5.5 inside it and typed a single line: "Claude can you edit these clips into a 10 minute vlog for X and YouTube". Almost an hour of footage across 14 files came out at roughly ten minutes, and Levels made only a few tweaks.
The fun part started when someone asked him how he edits. Levels went digging and found that the AI had invented the whole pipeline on its own. Speech was transcribed with mlx-whisper, meaning Whisper large-v3-turbo running on his MacBook Pro's GPU through Apple's MLX. To "see" the footage, Claude pulled frames every few seconds and stitched them into grids, like a photographer's contact sheets. SpeechBrain ECAPA, a voice matching model, separated his voice from other people's, OpenCV with the YuNet face detector made the thumbnails, and Python's Pillow library drew on-screen captions for X and for the first minute on YouTube.
The edit itself turned out to be a plain JSON file: which clip, start time, end time. Claude picked the pieces by itself, reading the transcripts and looking at the frame grids. Levels says he asked for none of this and had no idea that was how it worked.
You can try the same thing today: Claude Code, a folder of clips and a Mac that can run Whisper locally. Levels says that without AI he wouldn't vlog at all, because editing is slow and painful and hiring an editor would be even more hassle.
Freelance video editors just got a rival that works for a subscription and never argues about revisions.
Tools
Tools
Tools