Win the first 3 seconds, then let one shoot feed three platforms
Short-form video is not a smaller ad, it is a different language. AI can write the hooks and scripts and even auto-caption the cut. A real editing app (CapCut, capcut.com) and the platform apps do what AI cannot: film your face, remove watermarks, and publish.
Fastest path: one prompt, end to end
Paste this into any capable AI (ChatGPT, Claude, or Gemini). It turns one product into five ready-to-shoot clips.
🤖 AI prompt — paste into ChatGPT / Claude
You are a short-form video scriptwriter for a DTC brand. My product is [product] and my audience is [audience]. My one core benefit is [main benefit].
1. Write 5 hooks, each usable as one line of on-screen text a viewer can read in under 3 seconds. Use a different formula for each: contrarian claim, mistake warning, list tease, question, and before/after.
2. For each hook, write a matching 15-30 second script as 3-4 beats (one line each). Keep it native and spoken, not an ad read.
3. Give the on-screen text for each beat (captions, because most viewers watch on mute) and one clear CTA per clip.
4. Suggest ONE trending-audio type per clip (e.g. "upbeat trending sound") but tell me to pick the exact sound live inside TikTok, since only the app shows what is trending today.
Output exactly one markdown table:
| # | Hook (on-screen, <3s) | 15-30s script beats | On-screen text per beat | CTA |
If you cite any example clip as proof a hook works, open it and confirm it has 10,000+ views; put the view count and link next to it, or drop the example. Do not guess or estimate any number (view counts, trend data). Every figure must come from a tool you opened or a source you can cite; put the source next to it. If you cannot browse, leave that cell blank and tell me which app to open (TikTok Creative Center, ccc.tiktok.com) to fill it in.
Or do it in 6 steps
- Hook first, in on-screen text. Write the hook before you shoot. It must land as one readable line in the first 3 seconds. Per Teleprompter's 2026 guide, 71% of viewers decide within the first few seconds, and 50% of silent viewers rely on captions, so the hook has to be readable, not just spoken.
- Reuse a script template. Keep 3-4 proven skeletons (problem to fix, myth to bust, 3-tip list, before/after) and drop new products into them instead of writing from scratch.
- Batch 5 clips in one session. Run the prompt once, get five hook-and-script rows, then film all five back to back in the same setup and lighting. One session produces a week of content.
- Edit in CapCut (capcut.com). Auto-caption every clip, cut dead frames so it opens on movement, and keep each in the 31-60 second band Teleprompter names as optimal (33% of marketers) and the under-90-second band that holds about 50% of viewers.
- Post natively per platform. TikTok: add a trending sound picked live in-app plus native captions. Reels: export a clean file with no TikTok watermark, since Instagram's own Reels guidance says to avoid visibly recycled content from other apps. Shorts: write a searchable, evergreen title so it keeps surfacing in search.
- Adapt, do not re-film. Budget 5-10 minutes per platform to re-cut the hook, caption and audio from the same footage. Done = the same clip live on all three platforms, each feeling native.
Worked example
Say you film 5 core clips in one 60-minute session. Adapting each to 3 platforms at ~7 minutes per adaptation is about 105 minutes of editing, so roughly 15 posts from under 3 hours of total work.
Teleprompter's 2026 guide reports short-form earns about 2.5x the engagement of long-form, and videos under 90 seconds hold about half their viewers, so the shorter, native cuts are the ones worth batching.
Re-run the prompt weekly and keep the hooks that beat 50% completion.