Skip to content

gooseworks-ai/goose-skills / drafts/marketing/make-video-ads

make-video-ads

Verified

Use when the marketer says: make me a video ad, a 30-second ad for paid social, a Reel / TikTok / Shorts ad, a UGC-style video made with AI (no real creator), a product demo video, remake this ad I like for my brand, cut our best organic video into an ad, more like my winning video with new hooks, my video ad got rejected — fix it. Delivers a produced, ready-to-post video ad grounded in a researched angle: one message, a hook for the first two seconds, every claim traced to the brand's signed-off claims, no generated person posing as a customer, the exact words and the price approved before anything renders (one video, or one plus two hook variants), an honest review of the result, priced fixes only on a yes, and launch copy and notes for the ads team. Works in chat apps. Never publishes or spends ad budget. Not for real creators filming for you — work-with-creators (unsure? ask: AI-made or a real person); not for a phone-shot organic post — run-organic-social; not for static ads — make-static-ads.

references/script-craft.md

View on GitHub
# Script craft — paid social video

Load at Step 5. These are craft conventions, not performance promises; never quote CTR or hook-rate numbers as facts.

## 1. Beat structure

| Beat | 15s | 30s | 45–60s | Job |
|---|---|---|---|---|
| Hook | 0–2 | 0–2 | 0–3 | Stop the thumb. Picture, text and sound all land in the first second |
| Agitate / context | 2–4 | 2–6 | 3–10 | Make the problem specific, in the customer's words |
| Product + mechanism | 4–8 | 6–12 | 10–20 | What it is, why it works — approved facts only |
| Demo / proof | 8–12 | 12–21 | 20–40 | Show it in use; one proof point |
| Objection | — | merged 15–21 | 30–45 | Answer the top reason not to buy (from customer evidence) with an approved claim |
| CTA + offer | 12–15 | 26–30 | last 4–6 | One action, one reason to act now — said **and** shown |

Add a **re-hook** around 8–12s in anything 30s or longer: a visual change, a new question, "here's the part…".

## 2. Hooks

- **Triple hook:** picture (motion, a face, an unexpected image), text (≤ 8 words, readable in under a second) and sound (an effect or the first spoken words). Each must work alone — text for sound-off, sound for people looking away.
- **First frame = thumbnail.** No black frame, fade-in or logo intro.
- **The hook is the angle in miniature** and must pay off in the body. A clickbait hook the body ignores wins the first second and loses the sale.
- Specific beats generic: "3:07am" beats "late at night"; "people who rehearse conversations in bed" beats "overthinkers".
- No questions that assert a condition ("Can't sleep?", "Struggling with acne?") — rejected on Meta and a disease-claim risk.
- **Write 4–5 hooks before choosing one** (the engine takes 4–5 candidates). Pick by fit with the account: avoid hook families fatiguing in the performance read; prefer what is carrying the category in saved inspiration.

**Hook families** (name them; analyze-performance compares them): situation POV · identity callout ("for the people who…") · real-quote objection card · contrarian / "nobody talks about" · old-vs-new routine · ASMR / satisfying demo · founder confession · specific number (approved) · "watch this" demo · news ("new:").

## 3. Pacing and density

- **Voice-over: ≤ 2.3–2.7 words per second.** 15s ≈ 30–38 words; 30s ≈ 55–65 (pauses sell); 60s ≈ 130–150. Check the rendered duration — voices run fast or slow.
- **Cuts every 1.5–3s** on Reels/TikTok; about 1s in the hook. No static shot over ~4s without movement inside it.
- **On-screen text:** one idea per card, ≤ 8 words (≤ 12 for a quote card), held at least (words ÷ 3) + 0.5 seconds; at most two text elements at once (captions + one super).
- **Brand learnings override defaults and the examples** — a brand that said "spoken lines should flow as natural sentences, not fragments" gets flowing lines, whatever rhythm the worked example uses.

## 4. Sound-off

- Assume most feed views are muted. Read only the picture and on-screen columns: does the ad still land? If not, rewrite.
- **Burn in captions** for every spoken word, in brand style. Captions are claims too — they must match the approved words exactly.
- Music sits under the voice; voice intelligibility first.

## 5. Product and brand presence

- Product in frame by ~3s (a soft-focus cue in frame 1 counts), named by ~6–8s, back at the CTA.
- Shots where the label must be readable use the brand's **real product image** (`image_asset_id` from `media_list`); generated labels garble. Mark them `detail` / `readable_text` in a custom shot list.
- Hands-only shots feel native and avoid the fake-customer problem; they still need a finger and product-shape check.
- One message: one angle, one benefit, one proof, one objection. A second message halves both.

## 6. Native vs polished

- **Native / lo-fi** for cold prospecting on TikTok and Reels: phone framing, imperfect light, text-led, POV, hands-only, creator-style captions.
- **Polished** for retargeting, premium positioning, launches, YouTube in-stream, brand-safety-sensitive categories, most B2B on LinkedIn.
- "Native" is never achieved by faking a customer.
- How rough is rough enough is a taste call — surface it at the script gate.

## 7. CTA

- One verb plus one reason: "Try 30 nights — love it or your money back", not "Shop now".
- Said and shown, held ≥ 2.5s, not only on an end card most viewers never reach.
- The offer must exist in the product record or the marketer's words; never invented urgency.

## 8. Placements and versions

| Placement | Ratio | Length | Notes |
|---|---|---|---|
| Reels, TikTok, Shorts, Stories | 9:16 | 9–30s (TikTok up to 60s) | Safe zones: keep text and product out of the top ~14%, bottom ~25–35%, right ~15% (TikTok rail) |
| Meta / IG feed | 4:5 or 1:1 | 15–30s | Text cropped from a 9:16 must be re-checked |
| YouTube in-stream | 16:9 | 15–30s (skippable after 5s — the first 5s carry the brand) | Sound-on more common |
| LinkedIn | 1:1, 4:5 or 16:9 | 15–90s | Mostly muted; captions essential |

The engine makes one ratio and one length per video. Every extra ratio or length is a new video with its own price — plan the primary placement first; offer versions after.
add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 115kUpdated Claude CodeCodexVerifiedInstall