AI City Rendering Comparison Poster

Objective: To create a vertically side-by-side comparison poster showcasing two AI-generated aerial city renderings, arranged vertically, for use in [the context of the poster].{argument name="top model label" default="Banana pro"} and{argument name="bottom model label" default="GPT image 2"} Visual model comparison. Canvas: 3:4 portrait image, approximately 768 × 1024 pixels. The canvas is precisely divided into two equal horizontal panels with no borders or spacing between them. Layout: The top panel fills the upper half, and the bottom panel fills the lower half. Each panel presents a similar futuristic waterfront commercial district from a high, angled, bird's-eye view, set at a blues twilight. Two prominent large text labels are placed: the top label is located in the upper left corner and reads "Banana pro"; the bottom label is located in the upper left area of ​​the lower panel and reads "GPT image 2". Tall, square, compact white fonts with dark shadows or outlines are used, resembling retro templates or game typography. Top panel scene: A sophisticated architectural visualization of a modern urban development along a river or bay. It includes approximately 12 prominent glass office towers and mid-rise buildings arranged within a gridded park, featuring illuminated roof edges, landscaped green roofs, tree-lined roads, and warm streetlights. The top and lower right edges show water features, beyond which distant city buildings are visible. Orange and white light trails along the roads create a vibrant nighttime city effect. Bottom panel scene: A second version of the same waterfront area, slightly darker and more dramatic, featuring approximately 14 prominent glass towers and mid-rise buildings. Use cooler blue glass reflections, brighter window lighting, stronger road light trails, green rooftop gardens, and a river or bay curving along the top and bottom right edges. The composition should be distinctly similar to, but not identical to, the top panel to reflect two different AI model outputs from the same cue word. Visual style: Ultra-detailed aerial architectural rendering, realistic yet slightly idealized, cinematic twilight lighting, blue and cyan shadows, warm golden streetlights, crisp glass curtain walls, dense urban planning, dock/waterfront atmosphere, high contrast, sharp perspective, and no human presence. Constraints: Use two precise stacked contrast panels and two visible text labels. Labels should be large and clear, with white font and dark outlines/shadows. Do not add any logos, UI elements, watermarks, titles, or additional text.