Video Repurposing

Screen share

A screen-share recording turned into a long cut with chapters, shorts and a recap

Picture shape: 16:9, 9:16
Runs on Claude CodeCodexAntigravity

Turn one long screen-share recording (a webinar, workshop or online class, 30 minutes to 3 hours) into a full content pack: a clean long cut with chapters, one file per chapter, vertical shorts and a text recap.

Important: this is a skill for terminal AI coding agents (Claude Code, Codex) that runs on your own computer. It is not a prompt for the ChatGPT website. Check the requirements below before buying.

Why this look

A two-hour webinar sits on your drive because cutting it is a day's work. This look takes the recording (Zoom, Google Meet or Teams style, a shared screen with a small camera tile) and produces everything from one ingest, with ffmpeg only.

What you get

  1. A long clean version, 16:9 1080p, with silences, fillers and stutters cut, an SRT and YouTube chapters in cut time.
  2. One file per chapter (6-15 chapters, named in the speaker's words), split at exact frame boundaries.
  3. 5-10 vertical shorts, 45-90 seconds each: the slide on top, the speaker's tile in a bubble, a title card and highlight cards from what was said.
  4. A text recap (recap.md) with quotes copied word for word.

Every output ends with a short branded end card, which you can change to your own logo and message or turn off.

Privacy first

Attendee names, e-mails and private chat never appear on screen, in the SRT or in the recap. The taskbar, tab strip, chat panes and account menus are blurred on the source before the render, and the render stops until the source has a blur entry. A scan lists every place a name from the Zoom files could reach a viewer, and the recap check fails if one does.

How it works

Put the recording (and any Zoom files: transcript, chat, timeline, a separate camera file) in one folder, open your agent and type:

Làm trọn bộ nội dung từ bản ghi webinar này: bản dài có chương, 6 short và bài tóm tắt

The agent asks at most one message of questions: which outputs, how many shorts, the session start and end (the waiting time before the real start is trimmed), the cut level and music on the shorts. Then it transcribes, detects the screen layout, proposes cuts and chapters, renders and checks every file. The long cut of a 2-hour recording takes about 19-23 minutes.

What you need

  • Claude Code or Codex.
  • ffmpeg (an NVIDIA h264_nvenc encoder speeds it up) and Python 3.10 or newer.
  • OpenCV in a small Python environment, for layout detection, the speaker's head and the blur scan.
  • Edge or Chrome, used only to draw the cards.
  • About 30 GB free on the project drive.
  • SONIOX_API_KEY (or ELEVENLABS_API_KEY) for speech-to-text.

Install

Copy the skill folder into your skills folder:

~/.claude/skills/screen-share-repurpose/

For Codex, use ~/.agents/skills/screen-share-repurpose/.

Who it is for

One long screen-share recording (webinar, workshop, class) turned into a content pack.

Not a fit if you want

  • To edit short talking-head clips or footage with no screen share (use Fast cut).

Frequently asked questions

Can I use the videos commercially?

Videos you render are yours for any commercial use. You may not resell, redistribute or share the skill files.

More

Related looks