When to use it
You have raw footage and a timecoded transcript (a piece to camera, an interview, a podcast) and you need to turn it into an edit list: what to drop, what to keep, where to drop in a cutaway. The AI plays video editor. The result: a finished timecoded plan you can sit down and cut from, without rewatching for hours.
The prompt (copy and paste)
You are a video editor. Using a timecoded transcript, build an EDIT PLAN for the video.
FORMAT AND RUNTIME: <e.g. YouTube 8-10 min / Shorts 45 sec / Reels>
GOAL AND AUDIENCE: "<FILL IN what the video is about and who it's for>"
TIMECODED TRANSCRIPT:
<PASTE a transcript in the form [00:12] text... [00:48] text...>
Return a segment-by-segment script table, in the order of the final video:
| source timecode | keep/cut | what's said (brief) | role in the video (hook/point/example/CTA) | B-roll / on screen | title card/graphics | music/sound |
Rules:
1. Hook — the first 2-3 seconds: pick the strongest moment in the whole recording, even if it sits in the middle, and put it up front.
2. Cut ruthlessly: filler, repetition, verbal tics, long pauses, "ummm". Give timecodes for every cut.
3. Retention: flag where the pace sags and suggest a splice/cutaway/B-roll so the viewer doesn't drop off.
4. Where B-roll or on-screen text is needed, say SPECIFICALLY what to show.
5. At the end: an approximate final runtime after the cut, plus a "must cut" list.
Work only from the transcript provided and take timecodes from it. Do not invent lines that aren't there.
Why it works: a strict "timecode -> decision -> B-roll -> title card" table plus the "best moment goes in the hook" rule turns a transcript into an executable edit list rather than a summary.
Filled-in example
Format: YouTube, 8 min. Goal: a breakdown of "why 9 out of 10 startups die", for first-time founders. Transcript: 22 minutes of raw monologue with timecodes.
What the AI should return: the hook pulled from [14:30] "I burned 2 million and figured out one thing" (the strongest line, taken from the middle); a segment table with notes like "cut [03:10-04:05] — repetition", "B-roll: revenue chart over the money talk", "title card: 'Mistake #1'"; the sag at [09:00] covered with a cutaway; a bottom line of "~8:10 after the cut" and a "must cut" block listing the filler timecodes.
Variations
- For Shorts/Reels. "From this transcript, build an edit plan for a 45-second vertical clip: hook only + one point + a loop at the end."
- No timecodes. If the transcript has no timing, ask for cuts by meaning-based blocks with numbered labels.
- For a specific tool. "Lay the plan out as an edit list for Premiere/DaVinci: video / B-roll / title / music tracks."
Pro tips
- The single strongest move is pulling the best moment from anywhere in the footage into the opening seconds. Ask for it explicitly: by default the AI cuts "in order".
- Get the transcript automatically (Whisper or YouTube's own subtitles) — then the AI works from real timecodes instead of guessing.
- The plan is a map, not a verdict: trust your eyes in the edit. But the "must cut" list saves hours of reviewing raw footage.