When to use it
When you have a long video, podcast or webinar with a transcript and need to cut Shorts out of it. The role is a short-form producer. Result: 3–5 self-contained fragments with timecodes, a ready hook, a title and tags.
The prompt (copy and paste)
You are a short-form video producer. I will give you the TRANSCRIPT of a long video with timecodes.
TRANSCRIPT: "<PASTE THE TRANSCRIPT WITH TIMECODES>".
Find 3–5 fragments suitable for Shorts. For each one, reason it through and deliver:
- Start and end timecodes (a 20–60 sec clip, a complete thought).
- Why this moment will land (emotion / insight / conflict).
- A ready hook — the first line of the clip.
- A title and 3–5 tags.
Take only self-contained chunks that make sense without the context of the whole video.
Filled-in example
The transcript of a founder interview (45 min, with timecodes).
Expected AI answer: Fragment 1 — [12:30–13:10] the moment "we almost went bankrupt a week before the round"; it lands on emotion and drama; hook "Seven days before the round closed, we ran out of money"; title "How we survived one step from bankruptcy"; tags #startup #founder #business. Plus another 2–4 blocks like that, sorted by strength.
Variations
- Tuned to a clip niche. "Take only the argumentative moments / only the practical tips" — a consistent series.
- With subtitles. "For each fragment, output the subtitle text for the cut" — it speeds up editing.
- A thumbnail for each. Tie it to a cover concept: "suggest a thumbnail frame for the clip."
Pro tips
- Get the timecoded transcript from YouTube captions or Whisper — without timecodes the AI cannot give you accurate cut points.
- A long transcript (over 1–2 hours) may not fit in the context window — split it into parts or use a model with a bigger window.
- Self-containment is critical: a clip that only makes sense in the context of the whole video will flop as a Short. Hold that requirement firmly.