
Original document
GPT Image 2 · en
Scene:
A square moonlit storybook world built from crooked fences, leaning roofs, spiraled trees, pointed hills, and glowing orange windows against blue-violet night. The image should feel handcrafted and whimsical but also melancholic, with strong silhouettes and a theatrical sense of depth. Create a moonlit market lane lined with carved pumpkins, tilted awnings, and strings of dim lanterns between crooked houses.
Subject:
Include one central wanderer or small resident figure whose oversized eyes, slight body, and clear silhouette anchor the scene. The town itself should seem alive with curl, tilt, and delicate unease, as if every branch and roofline was bent by the same strange lullaby. A child figure in a stitched coat stands amid the pumpkins, face lit from below by one glowing lantern gourd.
Important details:
The visual law is gothic playfulness through spirals, cut-paper silhouettes, and warm-cool contrast. Preserve curling trunks, scalloped clouds, crooked pickets, and glowing window rectangles without referencing any existing characters or properties. Avoid cute modern merchandising polish or generic fantasy clutter. Use orange glows as punctuation in a blue-black world, keeping the market sparse and theatrical rather than festive crowd chaos.
Use case:
Gothic Fable Spiral Town single study 8/9.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no readable signs, no branded mascots, no realistic suburban details, no photoreal human anatomy, no daytime sunlight, no broken anatomy, no low-detail background, no accidental text fragments.How to use
- Copy the complete original prompt, including any parameters.
- Replace the subject, text and composition with your own requirements.
- Use a compatible external tool; check its current access conditions and review the result.
Before you start
- Check whether a reference image is required. A compatible model label does not mean this example was generated with that model.
- Keep names, figures and visible text accurate; results can vary between model versions.





