A Livestreamer's Claude Code Skill Turns a 2.5-Hour Stream Into 15 Captioned Clips and Shownotes
A livestreamer's Claude Code skill turned a 2 hour 26 minute stream into a transcript, chapters, shownotes and 15 captioned clips in about 2 hours. He says 15+ hours of editing became a few minutes of decisions.
Who this is about: Weekly livestream show on X (Brandon Davenport, former video editor)
What it takes
- Cost
- From $20/monthClaude Code: Included with Claude Pro ($20 a month billed monthly), Max, Team and Enterprise plans, but not the Free plan. You can also pay for API usage instead.Price checked Oct 24, 2026 on claude.com. Prices change; check before you buy.
- Setup time
- A few weeksOur estimate from the source, for one business.
- Skill needed
- Developer skillsYou build or run your own code, even if AI writes it.
Explain it simply
A creator who streams weekly built his own AI editing system instead of using a clipping app. One command starts it: it transcribes the stream, finds topics, suggests clips, cuts and captions them, and uploads them. He answers two questions and approves the clip list, and the rest runs while he does other work.
What was done
One command (a custom /show-wrap skill in the Claude Code CLI, plus local models) pulls the replay from X, normalises it and transcribes it locally with WhisperX.
Names are corrected through a ledger of fixes, checked against every name he has published before.
Five agents propose chapter boundaries and a boundary needs 4 of 5 to count; three agents propose clips and three score them.
He answers two questions at the start, then picks clips and signs off on a dry run of the cut plan. A push notification pings him when it needs him.
Approved clips are cut, captioned, loudness-normalised and checked, then uploaded to Google Drive, with picks queued on YouTube as private videos and a page pushed to his site.
How it went
For one 2 hour 26 minute stream, the run took about 2 hours and produced a corrected transcript, chapters, shownotes, 15 clips uploaded to Drive (5 queued on YouTube) and a live page. He says it saves 15+ hours a week.
- On that stream, 27 clip candidates were scored and he approved 15 of 26 shown. A final audit found 0 problems across 16 published links.
- He built a career in video editing, and the system encodes that judgment; it's bespoke, not a product.
- Promotional: he invites readers to DM him to build a similar system.
Prompt
Process the stream recording in [folder].
1. Transcribe it with word-level timestamps. Correct names using [names list]; log every correction.
2. Propose chapters with timestamps and a one-line title each.
3. Propose up to [20] clips. For each, quote the exact first and last words, give the timestamps, a title and why it works.
4. Stop and show me the list. Cut nothing until I approve.
5. After approval: cut each clip, add captions to vertical versions, normalise loudness, and upload to [Drive folder]. Queue my YouTube picks as private only.Setup notes
- Write down what a good clip looks like for your show before automating it.
- Keep an approval step on the clip list and keep uploads private until you publish them yourself.
Business fit
Who this helps, and how:
Related
- Evidence grade BA VR Review Channel Hands Video Editing to a Claude Code Agent and Approves Cuts by PhoneStrong evidence, using Claude Code, Gemini and Remotion
- Evidence grade BA Video Creator Hands Transcripts, Blog Drafts and Promo Posts to ClaudeStrong evidence, using Claude, Whisper (local transcription) and GitHub and Netlify previews
- Evidence grade BA Creator Cut LinkedIn Carousel Time From 2.5 Hours to Under 5 MinutesStrong evidence, using Claude Code and ILovePDF
- AI LinkedIn carouselsDepartment hub: AI for content, 18 examples
Similar tools
Other tools for this job, and how they differ. Chosen by our editors, never paid for.
“Strong evidence” (grade B) means: The person who did the work describes the setup and results in detail. How we grade. Spotted a mistake? Suggest a correction.