Creating Explainer Videos
Given a topic (e.g., "Why do we yawn when others yawn?"), produce the full 10-part output package below. Don't ask the user for more input — infer the best angle, write the script, and deliver a finished, production-ready package in one pass.
Progress:
- Find the unique angle (not just "what is X" — find the surprising lens)
- Write the hook (first 1-3 seconds)
- Draft script following HOOK → CONTEXT → EXPLANATION → TWIST → CONCLUSION
- Map every script line to a visual direction
- Add editing instructions with timestamps
- Add sound design
- Write captions with highlighted keywords
- Add CTA
- Fact-check and flag any unverified claims
- Run quality control pass, revise weak parts before presenting
Step details
1. Angle first. Before writing anything, ask: what's the one fact or reframe that makes this topic worth 45 seconds of someone's time? If the topic is "black holes," the angle might be "the thing you think is a hole is actually the opposite of empty."
2. Structure timing (adapt, don't force):
- Hook: 0–3s — surprising claim, odd visual, or question that creates a curiosity gap
- Context: 3–8s — minimum viable background
- Explanation: 8–25s — break into 2-3 simple beats, one idea per sentence
- Twist: 25–35s — the counterintuitive payoff
- Conclusion: 35–45s — memorable line, not a summary
3. Pair every line with a visual. No sentence should exist without a corresponding shot/graphic direction. If a visual is purely decorative (not explanatory), cut or replace it.
4. Timestamp everything in the editing guide — even rough estimates (0:00–0:03) are better than none.
Always output these 10 sections, in order:
- Video Concept — core idea + unique angle (2-3 sentences)
- Title — 3 options, each under 60 characters
- Hook — the exact opening line(s)
- Complete Script — timestamped voiceover, written to be read aloud
- Visual Storyboard — scene-by-scene, one row per script beat:
[Timestamp] | Voiceover line | Visual direction - Editing Guide — transitions, zooms, pacing, emphasis moments, timestamped
- Sound Design — music mood/genre per section, specific SFX cues with timestamps
- Captions — exact on-screen text, with bolded/highlighted words marked
- Final CTA — one natural, non-generic closing line
- Fact Check — list every factual claim made, mark as Verified / Common Knowledge / Needs Source; never fabricate statistics or quotes
Example 1: Input: "Explain why time moves slower near black holes" Output:
- Concept: Angle = "Your watch and a friend's watch near a black hole would disagree — and both would be right."
- Titles: "Time Is Lying To You Near Black Holes" / "The Clock Paradox Einstein Proved" / "Why Falling Into a Black Hole Freezes Time (Sort Of)"
- Hook: "If you watched your friend fall into a black hole, you'd never actually see them fall in."
- Script includes a beat comparing gravity to a trampoline stretching a grid (visual metaphor), explicitly flagged in Fact Check as "simplified analogy, not literal physics"
- Visual storyboard pairs each line with motion graphics of warped spacetime grids, not generic space stock footage
Example 2: Input: "Make a video about why Roman roads are still around" Output:
- Concept angle: "Rome didn't build roads to last 2000 years — they built them to not fail now. Durability was a side effect of obsessive short-term standards."
- Twist beat reveals the layered construction (gravel, concrete, fitted stone) as the "misunderstood part"
- Fact Check flags the exact road thickness/materials as "Needs Source — verify per region, construction varied"
- Open with the most interesting sentence in the entire script — don't build up to it.
- One idea per sentence. If a sentence needs "and," check if it should be two sentences.
- Use visual metaphors for abstract concepts (spacetime = stretched fabric, compound interest = snowball) but always label them as metaphors in the Fact Check.
- Keep the conclusion short — a single resonant line beats a recap.
- Write for the ear: contractions, short clauses, no subordinate-clause pileups.
- Vary shot/visual type at least every 3-6 seconds to sustain retention without being gimmicky.
- Don't open with "Did you know..." or "In this video, we'll explore..." — these kill the hook.
- Don't decorate with stock footage that doesn't explain anything — every visual earns its place by adding information.
- Don't pad runtime with restated information just to hit a target length.
- Don't invent statistics, quotes, or studies — if unsure, mark as "Needs Source" rather than stating it as fact.
- Don't let the twist be weaker than the hook — if the reveal under-delivers on the opening promise, rewrite the hook or find a stronger angle.
- Don't overload with transitions/effects; use pattern interrupts purposefully, not constantly.