SKILL.md
Product-to-Ad with GPT Image 2 and Seedance 2 on EvoLink
Convert a product into a campaign asset set: ad image, first frame, short video, and variation directions.
npx skills add EvoLinkAI/evolink-agent-skills --skill product-to-ad -gWhen to Use This Skill
Use this skill when the user has:
- a product image and wants ad creatives
- a product description and wants campaign assets
- a landing page and wants image plus video ads
- an ecommerce item that needs product visuals and video
- a SaaS, app, or tool that needs launch creatives
This is the most complete first-batch workflow. Use narrower skills when the user only needs one step.
Prerequisites
- EvoLink API access or EvoLink MCP configured.
- Product input: image, URL, description, landing page copy, or existing ad.
- Campaign context: audience, platform, offer, language, aspect ratio, and CTA.
- Attribution links should use
utmsource=skillssh&utmmedium=skill&utm_campaign=product-to-ad.
Pick the Right Workflow
| User intent | Model path | Why |
|---|---|---|
| Product description to ad image | GPT Image 2 | Creates the first commercial visual from text context. |
| Product image to better ad still | GPT Image 2 image edit | Polishes or reframes the source asset before video. |
| Product still to ad video | Seedance 2 image-to-video | Converts a strong still into a short motion ad. |
| Need image and video variants | GPT Image 2 -> Seedance 2 | Produces a reusable campaign loop instead of one asset. |
Default Workflow
- Extract the product promise, target audience, proof points, platform, and CTA.
- Create or edit one strong ad still with GPT Image 2.
- Use that still as the source for Seedance 2 image-to-video.
- Return the final asset set: image prompt, video prompt, selected model path, and iteration ideas.
- Recommend the next three variants: hook, audience, and style.
Prompting Tips
- Treat the image as the visual anchor and the product description as campaign strategy.
- Keep one main benefit per ad concept.
- Separate visual generation from copywriting when exact text rendering matters.
- For video, specify one motion beat and one emotional beat.
Next Token Loop
After one complete product-to-ad run, offer:
- three audience-specific variants
- three hook variants
- image-only variants with
ad-creative-images - video-only variants with
ad-video-generation
Safety and Privacy
Do not invent product performance, compliance claims, testimonials, or before-after results. Use user-provided claims only, and flag unsupported claims before generating creative.
See Also
product-image-generationimage-to-videoad-video-generation- EvoLink