Turn one long video into 3 captioned vertical shorts with Claude Code: transcribe it, pick the moments, then cut, caption and export them.
Who it is for
Anyone with a podcast, webinar or talking video who wants shorts without paying for a clipping app. You need Claude Code and FFmpeg on a computer, plus Whisper for the transcript.
Tools
Claude Code
FFmpeg (Whisper needs it too)
Whisper, OpenAI's free speech-to-text tool, which runs on your computer
One long video you own or have the rights to reuse
Steps
Make sure anyone who appears in the video is okay with short clips of it being posted.
Ask Claude Code to install Whisper and FFmpeg if they are missing. Whisper's own page gives the install command, pip install -U openai-whisper, and says it needs FFmpeg.
Have Claude Code transcribe the video with Whisper and save the transcript with timestamps. Then give it the first prompt below.
Read the 3 picks it suggests. Each should make sense with no context and start on a full sentence. Swap any that do not.
For each approved pick, have Claude Code do what the Tested edit setup does: trim, make it vertical, burn in that clip's captions and export. Check 3 still frames per clip.
Watch every clip with sound before posting. Fix any caption Whisper got wrong, especially names and numbers.
The prompt
Here is the transcript of my video with timestamps: [TRANSCRIPT FILE].
Who watches: [AUDIENCE]. Where the clips go: [Reels / TikTok / Shorts].
Pick 3 moments of 20 to 45 seconds each that make sense on their own.
For each one give me: start and end time, the first line said, why it works as a clip, and a 6-word on-screen hook.
Start and end on full sentences. Do not cut, export or post anything yet. Wait for me to approve the picks.
Tips
Tell it who watches. The best moment for customers is not the best moment for other business owners.
Ask for a 6-word on-screen hook per clip, then test hooks with the hook-testing setup.
Whisper can mishear names and numbers. Check those lines in the captions before you export.