skills/youtube-script/references/explainer.md

# YouTube Explainer / Education Scriptwriter

Write an original spoken-word script that teaches something. Patterns below come from the 20 most-viewed videos each of nine leading channels in the genre (~180 scripts). They describe craft, not personas — never copy any real creator's branded phrases or rituals; invent originals. The genre's core law: **the viewer is the protagonist** — the subject of the experiment, the ruler making the decision, the body being infected — never an audience member.

## Step 1 — Intake

Establish before writing (ask only for what's missing):

1. **The lane** — pick one; it sets the skeleton:
   - **Personality-led** (host on camera): builds, experiments, investigations. Stakes = the host's body, money, pride.
   - **VO essay** (voiceover over visuals/animation): science, economics, company/country stories. Highest word density.
   - **Inquiry-led** (host or VO): one question is the protagonist; the video is a detective case.
2. **The core question** — every explainer is contractually one question. If the user gives a topic ("quantum computers"), convert it to a question with tension ("why aren't they just faster computers?").
3. **The misconception** (if any) — what does the audience already believe that's wrong? A misconception flip is the strongest retention primer available.
4. **Target length** (~140–150 spoken words/min; VO essays run denser). If the target is ≤4 minutes, apply **Short-form mode** (below) instead of the long-form beat map.
5. **The concrete anchor** — one physical object, demo, number, or scene the explanation can keep returning to. If none exists, invent the analogy object in Step 3.

## Step 2 — Hook (offer 3 options, ≤30 seconds each)

Every hook does two jobs: plants a **concrete image** AND opens a **question the video must close**. Hooks are confident — hedging lives mid-video, never here. Draft three from different rows:

| Hook type | Formula |
|---|---|
| Artifact | "This is [object]. [Sharp number/superlative]. But [contradiction/question]." |
| Misconception flip | "You've heard [X]. That's wrong — so what's really going on?" |
| What-if | "What if [vivid scenario]? Surely [catastrophe]? Well… almost." |
| Impossible object | "[Record-holding thing] exists — and it breaks a rule you trust." |
| In-medias-res | "This is me [inside the climax]. To see how I got here, go back [N months]." |
| Body-count anecdote | "In [year], in [place], [specific losses]. Nobody knew why." |
| Dare | "You can't do [X] / this is your last chance to look away." |

Optional accelerants: a 15-second news-clip montage right after the hook (certifies the topic is real and contested); sharp numbers over round ones always ("$1,444", not "about $1,400").

## Step 3 — Contract, then structure

**Contract (0:30–1:30)**: state the stakes in the viewer's own life, then pose the explicit question list — ideally exactly 3 questions, answered in order. The roadmap is itself the retention device: each question is an open loop with a named payoff.

**Beat map (% of runtime):**
- **12–30% Primer** — teach the minimum mechanism through ONE anchor analogy (see analogy formula below).
- **30–70% Escalation ladder** — the same mechanism at increasing scale or through successive cases. Each rung reuses the previous rung's vocabulary. A "but" pivot at least every 90 seconds ("but here's the weird part…").
- **60–75% The turn** — the reversal the hook promised. Consider a **fake ending**: script the wrong conclusion fully, even "The End." — then demolish it.
- **85–95% Zoom-out** — reframe to humanity, time, or the viewer's identity. The end gets the *meaning* payoff; the biggest *factual* payoff already landed at ~70%.
- **Final 5%** — optional one-line outro (the next question, or a plain subscribe line); or simply end on the zoom-out.

**Lane variants:** personality-led = quest chronology (dream → counted failed attempts → test day → payoff), with the dense science block quarantined AFTER emotional buy-in and explicitly bargained for ("give me two minutes"). Inquiry-led = evidence in the order the investigator found it, with checkpoint questions ("but hang on — how do we know that?"). VO essay = causal chain, one domino per paragraph, twist at domino five.

## Step 4 — The analogy formula (mandatory for any abstract concept)

1. **Flag it**: "Think of it like…" / "Imagine…"
2. **Run the object's own logic** for 2–4 sentences, zero jargon (the event horizon is a river ending in a waterfall — stay in the river: current, swimmer, point of no return).
3. **Return with the technical name attached.**
4. **Reuse it as shorthand** for the rest of the script.
5. **Admit where it breaks** — the admission buys trust for the next analogy.

Supporting rules: personify invisible agents (cells "yell bloody murder"); give every large number a household ruler (Everests, Titanics, wallets, marathons); calibrate sensations by naming a familiar one then correcting it.

## Step 5 — Register & retention rules

- **Second person by default**; first person only as fellow learner, one step ahead at most. Confess ignorance freely — it licenses the explanation.
- Present tense, even for history (set the date, then narrate live). Contractions everywhere.
- Rhythm: a paragraph of ramping clauses lands on a **blunt 3–6 word verdict sentence**. The short sentence is the aha.
- **Jargon protocol: show first, name second, defuse immediately** — demo before term, then a domestic image or joke within one sentence of any technical word.
- No 90-second stretch without a reveal, reversal, or laugh.
- Every IOU is paid: "more on that later" is a promise with a due date inside the same act. Track and close every loop on screen.
- Reference only setups the script actually planted: never write "as I promised" or "remember X from earlier" unless the earlier line exists in this script — a false callback is an instant credibility break.
- When a claim can be tested, stage the test instead of citing it.
- Hedge in the middle plainly ("the truth is, we don't know"), never in the hook.
- Frame the viewer's misconception as reasonable before correcting it — the viewer must never feel stupid.

## Step 6 — Ending & outro

Do not invent sponsor reads, brand integrations, product plugs, or other monetization beats. Include a specific monetization integration only when the user explicitly requests it, outside the ladder, turn, and zoom-out. Otherwise the zoom-out IS the ending: reframe to humanity, time, or the viewer's identity, then stop. Optionally extend the curiosity with one plain closing line — the next question, or an invitation to follow it — never begging, never more than a sentence.

## Step 7 — Anti-patterns (hard avoid list)

- No definition-first openings — definitions arrive only after a tension exists that they resolve.
- No "today we're going to talk about…" — show the topic, don't announce it.
- No lecture register: no "as we can see", "it is important to note", "in conclusion".
- No unbroken abstraction beyond ~30 seconds without a physical anchor, a ruler-ed number, or a joke.
- No unpaid IOUs, no moralizing before mechanism, no round numbers where sharp ones exist.
- No invented monetization beats; only include a specific integration when explicitly requested. No press-release register anywhere.

## Short-form mode (target ≤ 4 minutes)

The beat maps above were measured on long-form; under ~4 minutes the physics change. Apply these overrides and keep the register rules (viewer-as-protagonist, show-then-name, no naked numbers):

- Budget: ~150 wpm → a 3-minute video is ~420–470 spoken words.
- **One question, one analogy, one turn.** The misconception flip is the ideal short hook: first line states the wrong belief, second line breaks it.
- Skip the question-list contract entirely — the single question IS the contract.
- Structure collapses to: hook (0–10%) → the one anchor analogy carrying the mechanism (10–60%) → the turn / "but here's the weird part" (60–85%) → payoff + one-sentence zoom-out, hard stop.
- **One loop only**, opened in the first line, paid at the end. No IOUs, no fake ending, no escalation ladder beyond one rung.
- Timestamps every 15–30 seconds; retention map 3–5 lines; outro one plain line or cut.
- If vertical (Shorts): the concrete image must be on screen in the first 2 seconds and the hook must work muted; consider a loop-close.

## Output format

```
# [Working title] — [lane], ~[N] min

## Hook options (pick one)
[3 hooks, each 40–80 words, labeled by type]

## Script
[Timestamped sections: 【0:00】…]
[VO as flowing spoken prose; verdict sentences on their own lines]
[Cues in brackets: [DEMO: …] [ANIMATION: …] [B-ROLL: …] [ON-SCREEN: 1,444]]

## Retention map
[One line per question-contract item / IOU / "but" pivot / fake ending, with timestamp]
```

Claims about the script belong in the script: no self-reported pacing or compliance boasts in titles or headers — the retention map is the only meta-commentary, and it lists devices, not grades.

For full pattern detail (9-hook taxonomy with illustrative examples, per-lane beat maps, payoff cadences), read `references/explainer-patterns.md`.