How Brooke Lacey Uses Claude Code to Edit Short Videos

Claire Vo interviews creator Brooke Lacey about using Claude Code and local media tools to edit short videos and develop content ideas. Lacey describes her own workflow and promotes her AI-learning app; the speed and results are examples, not guarantees.

Lacey says precise jump cuts and comedic timing used to make editing laborious. Years of making videos helped her explain the look and rhythm she wanted.

She describes recording a long phone take, including spoken instructions to remove failed attempts, and transferring it to her computer. She then asks the coding agent to work with the local file.

The workflow uses familiar tools including FFmpeg, Whisper and Python. Lacey says the raw file stays on her computer for this editing step; what information leaves the device depends on the tools, prompts and services actually configured.

In the demonstration, the agent produces a vertical clip with cuts, captions and changes in framing. The participants describe the run as taking seconds, but the timing is from one demonstration and excludes the work of developing the workflow.

Lacey walks through silence detection, word timing, face tracking and a text-based cut list containing start and end times and framing choices. The output can also include app screenshots and captions.

She emphasizes that she already understood short-form editing and used that experience to specify cuts, hooks and cadence. She says capturing what makes a joke work remains more difficult than automating mechanical edits.

For idea generation, she places copies of her own videos in Google Drive and asks the agent to examine comments on a selected TikTok post. It combines the video and audience discussion to suggest new topics and scripts.

Lacey gives feedback about her delivery and asks the agent to look for patterns in her hooks and timing. Vo summarizes the workflow as using examples and human feedback to refine both ideas and edits.

Lacey says she posts roughly weekly and is exploring live streams and ads. Her audience and posting frequency reflect her existing experience and following rather than a recipe for guaranteed growth.

When the agent produces an unsatisfactory result, she directs it back to project instructions and status files. She describes maintaining separate context for projects and revising those instructions when she finds a recurring problem.