/generating-image-ads

name: generating-image-ads
description: |
  Turns a product photo or URL into a static image ad — one image, or a set. Reads the real product first and holds it identical across every frame and ratio. Triggers: "image ad", "make an ad", "static ad", "promo graphic", "offer graphic", "sale creative", "social ad", "banner ad", "ad creative", "before/after ad", "testimonial ad". Not for generating product photography.
license: Apache-2.0
metadata:
  version: "0.1.0"
  category: creative
  summary: "Produces a scroll-stopping image ads for your product — one image, or a set in every placement ratio your feed needs. It keeps your real product and your brand identity perfectly consistent across every frame, and shapes the offer and claims you give it into a well designed ad, following industry best practices."

Image ads

Turn a product into finished static ad(s).

Workflow

Step 1: Read what you have

• **A product image or URL is available** → hand it to `analyzing-products`, and take four things

back: what the product is; **how a person physically uses it** — sprayed, pumped, worn, etc.; which parts open or move, and in what order; and the details that have to stay identical wherever it appears.

• **Look at the image before you describe it.** Take the product only from what the image shows. Run `image_analysis` and ask it for everything this ad needs that the product facts leave out — how the surfaces take light, and whatever else the frame you end up building will have to describe.
• **Product not supplied** → don't guess at the product. It becomes the first thing Step 2 asks for.

Step 2: Interview

**Skip this entirely when the brief already makes the ad obvious** — a clear product, a clear message, and nothing load-bearing missing. The interview exists to close real gaps, not to confirm what you were already told.

Otherwise ask once, bundled into a single message. Always leave a free-text way out so the user can answer off-menu. Anything this skill already settles is not a question.

**Do ask, when the answer is genuinely missing:**

| Ask | When | | --- | --- | | **The product** — what it is, and whatever else identifies it: packaging, colour, distinguishing features | No image and no URL. Offer to wait for an upload first; a photographed product beats a described one. | | **Brand guidelines, if they have any** — the color palette, the art direction they shoot to, anything they never put in frame, the typeface where text is drawn, and the register they want the frame to read at | The brief mentions none and the packaging doesn't imply them. Treat as optional: plenty of brands have none, so take what they give and move on. | | **The offer and the one claim** — a discount, a launch, a benefit, the words they can stand behind | The ad has to say something and the brief doesn't say what. Only claims the user supplies are used, and used as written. | | **Where it runs** — the placement and the ratios | The brief doesn't say, and the placement decides the crop and the safe zones. |

**When the user waives questions** and the product is already in hand, go straight to generating. Take a single image at `1:1`, a product-hero direction, and one benefit-led line drawn from the product facts. Say in a sentence which settings you took, then keep going. It is a statement, not a checkpoint.

A missing product image is the one thing still worth raising.

Step 3: Find the idea

Read `references/ad-craft.md` first. Work in the discipline of the single-minded proposition: every ad resolves one tension between a consumer truth and a product truth.

• **The insight** — a human truth about how a person uses or feels about this kind of product, something true enough that it feels slightly uncomfortable to say out loud. Let that truth create tension.
• **The idea** — resolve the tension into one sentence describing what the ad is about. If you cannot say it in one sentence, you have two ads, not one. Where the brief carries a promotion — a price, a discount, a launch — weave it into the idea as a real element, not bolted on afterwards.
• **The treatment** — only then decide how the idea becomes a single still frame: the mood, the register, the visual approach, chosen because it serves this idea. Treatment follows idea, never the other way round. It may take a familiar shape — a hero frame, a lifestyle scene, a before/after, an offer — but it is chosen to serve the idea, not picked first.

For a set, each ad works from a different **insight** — a different human truth, a different angle into the product — not a different visual treatment of the same insight. If two ads share an insight, you have shipped one ad twice. Nothing is generated here.

Step 4: Write the copy

Write the copy last — say only what the picture leaves unsaid. From the claims the user supplied and no others:

• **A headline** — the dominant text element, always present. Could be one word, a short sentence, or a wordmark-style brand line — whatever the ad needs.
• **Support** — an optional second tier of text. Use only when the headline alone leaves a real question unanswered. Most ads do not need this.
• **A call to action** — an optional 2–4 word direct-response action, rendered as a **button** (a filled shape with the action inside, set in the accent colour — never floating text).

Every drawn string is one the brand can stand behind. **Don't invent a claim, a statistic, a rating or a badge** to fill the layout.

Step 5: Write the product description

Write one description of the product, and reuse it unchanged in every prompt you send.

**How much to write depends on whether a photo is going with the call.** With a reference attached, the photo carries identity and the description covers only what it cannot: the angle you want, the state the moving parts are in, the action being performed. With no reference, the description is all the model has, so it carries the product in full from the list below.