SQL-style prompt gets GPT Image 2.5 to render foodie cities as 3D type posters
The author used a pseudo-SQL prompt to have GPT Image 2.5 turn four famous foodie cities into 3D typography and still-life posters.
I asked GPT Image 2.5 to visualize 4 famous foodie cities around the world. Prompt: WITH identity AS ( SELECT derive_display_name([PLACE_OR_CULTURE]) AS display_name, infer_signature_foods([PLACE_OR_CULTURE]) AS foods, infer_signature_materials([PLACE_OR_CULTURE]) AS materials, infer_landmarks([PLACE_OR_CULTURE]) AS landmarks, infer_objects([PLACE_OR_CULTURE]) AS objects, infer_palette([PLACE_OR_CULTURE]) AS palette, infer_three_values([PLACE_OR_CULTURE]) AS values, infer_keyword_stack([PLACE_OR_CULTURE], 4) AS keywords ), glyph_assignment AS ( SELECT glyph, argmax( source_item, shape_match(glyph, source_item) * cultural_relevance(source_item) * visual_uniqueness(source_item) ) AS source_material FROM letters(display_name) CROSS JOIN cultural_pool(foods, materials, objects) ) SELECT render FROM editorial_travel_stilllife_archive WHERE hero_typography = build_3d_word( display_name, material_per_glyph = glyph_assignment.source_material ) AND foreground = infer_culinary_stilllife([PLACE_OR_CULTURE]) AND background = infer_soft_focus_architecture([PLACE_OR_CULTURE]) AND base = 'premium sculptural plinth' AND base_caption = join(values, ' • ') AND side_stack = keywords AND lighting = 'warm sunlit premium editorial' AND styling = 'travel magazine × culinary still life × crafted typography' ORDER BY cultural_specificity DESC, glyph_legibility DESC, material_variety DESC, tactile_realism DESC, composition_balance DESC LIMIT 1;