Skip to content

gooseworks-ai/goose-skills / skills/video/write-video-ad-script

write-video-ad-script

Verified

Write or revise a short-form ad from a sourced angle and reference. Plan connected words, product proof and sound; expose production gaps before script review.

references/scene-and-hook.md

View on GitHub
# Human version

**Summary.** Plan the scene and the opening the viewer receives before writing dialogue.
An interview can begin after the approach, but its first image and words must establish
enough context to understand the exchange. Build curiosity about an answer or outcome;
do not make the viewer guess what the conversation is about. Read the lines aloud only
after the scene, question and payoff make sense.

The failed Som and GooseWorks drafts opened with answers to questions the audience never
heard. Private notes about an opted-in participant did not supply that missing context.
This guide changes how the next concept is written and reviewed. It does not establish
that the next script will pass human review or perform as an ad.

---

# Agent version

## Research findings and their limits

Research and opening-only diagnostic: 2026-10-03.

| Primary source | What it supports | What it does not establish |
| --- | --- | --- |
| [NBCU Academy: street interviews](https://nbcuacademy.com/street-interview/) | Journalists explain their purpose, connect the approach to the person or setting, and listen for the next question. | A commercial ad template, conversion performance or a rule that every edited clip must retain the approach. |
| [Hannah Williams: Salary Transparent Street](https://www.creatorspotlight.com/p/hannah-williams-salary-transparent-street) | The creator describes a clear purpose and permission before filming a short interview; visible branding helps identify the series. | That this approach appears in every published edit, or that these techniques sell either test brand. |
| [YouTube: Jenny Hoyos and Todd Sherman on Shorts](https://blog.youtube/creator-and-artist-stories/youtube-shorts-deep-dive/) | The creator emphasizes the opening and a short story that delivers on the interest it creates. | A universal need for shock, a fixed hook phrase, or a guaranteed result. |
| [TikTok Creative Codes](https://ads.tiktok.com/business/en-US/creative-codes) | Plan hook, body and close together; picture, movement, text and sound can carry the opening. | That adding a provocative sentence makes an otherwise confusing scene work. |
| [Faber Academy: dialogue](https://faberacademy.com/reading-room/writing-exercises/how-to-write-better-dialogue/) | Intentions, relationships and the moment shape what people say. Read dialogue aloud; avoid characters explaining facts solely for the reader. | A street-ad example or instructions to add fictional conflict to every interview. |

These are technique sources. **They do not replace the complete commercial reference**
required by the `render-street-interview` skill. No human script authorship, advertising
performance or accepted benchmark score is inferred from a real person appearing on camera.

### One observed published opening

Source: [Salary Transparent Street: Chicago PhD student](https://www.snapchat.com/@sts.ca/spotlight/W7_EDlXWTBiXAEEniNoMPwAAYbnRlZGNmd2Z0AZy5cBqDAZy5cBoBAAAAAQ).
Inspection: selected paused video frames at 0.000, 2.223 and 16.140 seconds, plus the
published opening caption; not continuous viewing or direct audio listening. Full player
duration: approximately 68.7 seconds. Sponsorship, recruitment and script authorship are
unknown. This record covers an opening technique, not a complete commercial interaction.

- At 0.000 seconds, both people and the extended microphone are visible. The video has
  a Chicago location label. Its first caption begins the question about the person's work.
- The published opening caption gives: “What do you do for a living? I'm a PhD student.”
- At 2.223 seconds, the interviewer is visible again and the caption reads “Cool,”.
- At 16.140 seconds, the participant has a PhD-student role label and a salary amount.

**Inference for our workflow:** the image establishes an interview and the complete
question establishes its subject before the answer. The edited opening needs no filmed
greeting to be understandable. A microphone alone would not explain an answer about an
unknown person's nighttime thoughts or an unknown team's advertising work.

## Plan two starts, not one

Record the **encounter start**: who is approached, where, why this person could answer,
the invitation or task, and what they know before filming. Mark this as an authored plan;
keep an observed source's unseen recruitment unknown.

Separately record the **viewer opening**: the literal first image, audible words, visible
action and any exact text that will actually appear in the cut. Outside captions, the
brand brief, a file's title and private character notes are not visible context.

Starting after the approach is allowed. Starting after the explanation is allowed only
when the retained words, action or readable on-screen text supply the necessary meaning.
A reply-first opening needs an identifiable subject or task. Do not insist on a greeting,
an answer-first edit, a mystery or a startling statement for every brand.

## Write the short story before its sentences

Save this compact record alongside the situation brief:

| Decision | Required output |
| --- | --- |
| Encounter | Location, participant role, interviewer purpose and authored invitation/task |
| First 0–3 seconds | Planned time, exact visible action/frame, audible words and rendered text, if any |
| Viewer understanding | What someone new can understand from those inputs alone; note any necessary missing information |
| Viewer interest | The specific answer, choice or outcome the opening makes them want to see |
| Development | What the reply or action adds, and how it changes the next question or action |
| Payoff and brand role | What is answered, revealed or made useful by the end; the supported role of this offering |
| Production fit | Time for the actual action and speech; unavailable assets or unsupported execution |

Do not turn this into compulsory problem-question-pitch slots. A product definition or
logo is not automatically a story payoff. If the intended discovery needs more time than
the recipe permits, choose a simpler encounter or flag a different production route.
Do not remove the necessary question to fit the clock.

## Write speech for the moment

Write what this person would say to the other person, given what both currently know.
Read it aloud. Ask whether the reply answers the actual question and whether an ordinary
listener understands its intended meaning. A sentence can be grammatical yet vague,
written-sounding or inappropriate for this moment. Fix the intent and referents before
adding contractions or trimming words. Do not insert filler, fake laughter or hesitation
to pass a human-speech check.

## Review without the explanatory brief first

Give a fresh reviewer only the literal opening's picture/action description, audible
words and rendered text. Withhold the campaign brief, source explanation, character
intentions, participation notes and ending. Ask the reviewer to state what is happening,
what the subject is, and what answer or outcome they expect next. They must identify the
visible/audible evidence for each statement, and flag any invented context.

If the opening needs private information to become intelligible, rebuild it before
scoring the full ad. Curiosity about the answer is useful; confusion about the question
is a defect. This check is an additional diagnostic, not a score that changes the frozen
Jev rubric or evidence of independent human acceptance.

Then provide the complete scene, dialogue, brief and brand facts for review of turn logic,
spoken wording, buyer relevance, claims and payoff. Keep conflicting reviews visible.
Only the manager's review establishes acceptance in this experiment.

### Diagnostic result: basic comprehension is too weak an acceptance test

A fresh reviewer, given only the two human-rejected first utterances and their planned
frames, still found them plausible and largely understandable. It identified the missing
fatigue/planning connection and unexplained team/terminology, but did not reproduce the
manager's rejection. It also understood the observed complete occupational question.
This was one same-family reviewer with text descriptions, not performed video.

Keep that disagreement visible. The opening check can expose unsupported context; it
cannot approve natural speech, an interesting story or an ad. Do not equate "a viewer
could infer the topic" with "this is a clear, compelling opening." Test the whole story
and keep human-rejected examples rejected even when a new model review rationalizes them.
receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementationStars: 322kUpdated Claude CodeCodexVerifiedInstall
requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirementsStars: 322kUpdated Claude CodeCodexVerifiedInstall
using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallbackStars: 322kUpdated Claude CodeCodexVerifiedInstall
verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions alwaysStars: 322kUpdated Claude CodeCodexVerifiedInstall