Atelier Zero — Image Generation Prompt Pack
This pack is consumed by the open-design-landing skill. Every page-level
image is rendered with gpt-image-fal (preferred) or gpt-image-azure.
The pack has three layers:
- Style anchor — the long block that tells the model what universe we are in. Always prepend to every prompt.
- Variable slots — the per-render content (subject, motifs, accent, page type).
- Per-slot variants — explicit composition templates for hero, about, capabilities, method tiles, lab cards, work cards, testimonial, and CTA.
Render at 1024×1024 minimum for square slots (hero / about / capabilities
/ testimonial / cta), 816×816 for the four method tiles, and 768×1024 for
portrait slots (lab cards, featured work). Authoritative per-slot
dimensions and aspect ratios live in image-manifest.json — treat that
file as the source of truth. Save as PNG to assets/<slot>.png.
1. Style anchor (always prepend)
Use case: ads-marketing
Asset type: editorial website hero / creative studio landing page visual
Primary request: Generate a refined editorial web page composition in the
same visual language as a high-end creat…
