Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

Brooke Lacey air-dropped a raw phone video to her Mac, then had Claude Code use FFmpeg, Whisper, and Python to cut a vertical short locally. She stresses the footage never uploaded anywhere; those ordinary desktop tools simply ran on files already sitting on the computer.

Related Insights

To scale short-form video, HubSpot created an automated clipping engine in a six-week sprint. It uses AI (Claude) to identify clips and an API (Descript) to edit them, leaving the final creative approval to a human. This "human-in-the-loop" system massively increased output without sacrificing quality.

Move beyond simple AI editors by using a plugin like Remotion that connects to a large language model such as Claude. This allows you to provide editing instructions in plain English, describing the desired B-roll, captions, and cuts, letting the AI execute your creative vision.

Marco Casalaina used natural language to instruct an AI agent to re-encode a 1.7GB video file using the powerful but complex command-line tool FFmpeg. The AI handled the specific command generation, reducing the file to 13MB. This makes highly technical tools accessible for tasks like file manipulation without requiring deep expertise.

Brooke Lacey says Claude tracked her face every frame, assigned each cut a wide, mid, or close shot, and placed app screenshots on the empty wall by measuring her right edge per cut. Forty of 441 frames carried the app screen from the rendered footage.

AI can automate tedious video editing tasks like finding the best takes from raw footage and adding blurs. For solo creators, this also offloads the emotionally draining process of watching and critiquing their own content. By delegating the 'critique' to a neutral AI, it lowers the psychological barrier to content creation.

While on a bike ride, Jason used his phone to remotely command his AI to find a video file on his laptop, re-export it with requested subtitle changes, post it to Slack, and then autonomously handle subsequent revision rounds based on team feedback.

Brooke Lacey explains that silence detection alone cannot place a clean cut. Claude pairs it with Whisper's word timing so each cut lands right after the spoken word ends, because the audio tail finishes sooner than people expect in natural speech. That tiny gap is what makes the edit feel abrupt.

Video editing can be automated without technical skill by using an AI as an orchestrator. By giving Claude access to a plugin like Remotion, you can use natural language prompts to direct it to access raw video files and perform complex edits, turning a creative brief into a near-finished video automatically.

Instead of writing detailed prompts, record a screen-capture video with a tool like Supercut while verbalizing desired changes. Its AI can accurately parse long videos and map transcripts to frames, providing enough context for a model like Claude to build the feature directly with high accuracy.

Brooke Lacey says the whole edit is a TSV file: each row holds a start, end, shot size, and label. Because that text file defines every cut rather than a timeline, the same command can generate 30-second, 60-second, and landscape versions from the same take.

Claude Code edits video locally using FFmpeg and Whisper | RiffOn