Adapt Brooke Lacey’s Claude Code workflow to edit phone footage with FFmpeg, Whisper, and Python. Use a TSV cut list to review timing, then render a vertical video with captions and graphics.
with Claire Vo
Okay, my dude, I'm ready to show the audience the magic.
A rendered, 15-60 second vertical video with tight jump cuts, one-word-at-a-time captions, and graphics placed in the frame. The process also generates a TSV (Tab-Separated Values) file that defines the entire edit.
Claude for coding tasks
FFmpeg is a collection of tools for recording, converting, filtering, and streaming audio and video.
Whisper is OpenAI’s speech-recognition model, used here for transcription and word timing.
Python is a programming language used here to coordinate local video-processing tasks.
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
5 steps
Film a continuous take on your phone. Don't worry about mistakes; you can repeat lines and even say 'cut that out' as a cue. Transfer the single video file to the computer where you will run the workflow.
When filming, intentionally leave empty space in the frame, like a blank wall next to you. This creates a usable area for inserting graphics later.
With Claude Code, FFmpeg, Whisper, and Python installed and available, ask Claude Code to analyze a short local clip. The example prompt below is an adaptation, not Brooke’s original script. Her demo used a prepared editing setup.
Using FFmpeg, Whisper, and Python, analyze the video at `[path/to/your/video.mov]`. Your goal is to create a 15-second vertical ad. First, identify all speech, use word-level timing to find precise cut points, and track my face to determine shot composition.
The guest used a simple command to trigger a pre-built script. For a new setup, a detailed, multi-step prompt is necessary to guide the process.
The system will produce a TSV (Tab-Separated Values) file where each row defines a cut with a start time, end time, and shot size (e.g., 'wide', 'mid', 'close'). This file is the entire edit plan.
From your analysis, generate a TSV file of the best cuts for the target duration. Each row should contain a start time, end time, a shot size, and a label. Prioritize takes with the best cadence and exclude segments where I said 'cut that out'.
This TSV file is your primary point of control. Before rendering, open this file in a text editor to check the timestamps. Adjusting a number here is much faster than re-editing a video project.
Instruct Claude to add visual overlays and captions, specifying their appearance and timing.
Now, add overlays. Take a screenshot of my app from the simulator and place it on the empty wall to my right, sized to fit the space in each cut. Also, generate one-word-at-a-time captions in white text with a black outline for the entire clip.
Specificity is key. The system was told to measure the empty space in each cut to size the graphic correctly, a more robust instruction than just 'put it on the right'.
With the edit plan finalized in the TSV and graphic instructions provided, command Claude to assemble and render the final video file.
Using the TSV cut list and the instructions for graphics and captions, render the final video using FFmpeg. Ensure the final output is a 15-second vertical video at 30 FPS with zero audio/video offset.
After rendering, watch the entire file to check word boundaries, caption placement, graphic placement, and audio/video sync. Brooke described additional formats; test each export rather than assuming it works.
Turn an idea into a PRD, user stories, and a plan.
How to recover when the loop fails and where human judgment helps.
Let routine work keep moving.
A step-by-step guide to signing up for Meta's Muse API, installing the Muse Code command-line tool, and running it headlessly with an external agent like Claude Code for easier testing and benchmarking.
Benchmark an agentic AI model's ability to resolve complex customer billing issues. This workflow tests the AI's ability to investigate tickets, handle complex logic, execute changes in an admin app, and communicate with the customer.
Use your existing TikTok video library and their associated comment threads to generate highly relevant new content ideas and scripts. This workflow pairs audience feedback directly with the specific video performance that prompted it, leading to targeted and engaging new topics.
Join 250,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.
