paper-illustration

Produces the prompt, not the image. The deliverable is a text prompt the user pastes into an image model together with the photo as a reference image. One photo in, one prompt out; several photos in, several separate prompts out — never merge photos into one illustration.

This is the illustration half of editorial-poster lifted onto its own canvas. The default canvas is a full 3:4 portrait, treated as the cover of a small independent art book; the subject stays small (10–20% of the canvas) so the empty paper does the work. The template fits a 2000-character input box (verified with GPT Image 2.5 Pro) with about 300 characters spare.

Step 1 — Collect input

• Photo(s) — attached to the message, or described in words if the user cannot attach.
• Text to include (optional) — a title, place, year, or short phrase. If nothing is given, use none.
• Aspect ratio (optional) — default 3:4 portrait. Only change it if the user asks (3:2 landscape reproduces the poster's bottom half exactly).

Do not ask for anything else.

Step 2 — Emit the prompt

Output the template below verbatim, replacing only <text or none> (and the aspect ratio in the first line under FORMAT, when asked). Put each prompt in its own fenced code block. For multiple photos, label the blocks Photo 1, Photo 2, … in the order received.