GPT 템플릿: GOAT 선수들을 위한 SQL 스타일 전기 몽타주 포스터

Prompts like this show how much better certain image models are. In this case, it's GPT Image vs. Reve vs. Nano Banana vs Seedream 5.0 Pro. Prompt: 2x2 grid, 9:16 for GOAT athletes<instructions> SELECT poster_image FROM legacy_montage_collage_engine WHERE subject = '[TOPIC]' AND image_type = 'celebratory tribute / career-history montage' AND composition = ARRAY[ 'vertical poster', 'central heroic anchor near lower middle', 'large emotional portrait or iconic symbol in center', 'dense surrounding cloud of smaller moments', 'achievement objects clustered near center', 'environment / home base along bottom edge', 'fan signs, banners, handwritten notes, dates, slogans' ] AND scene_logic = ARRAY[ 'show multiple eras of the same topic', 'show many emotional states', 'show triumph, process, struggle, celebration, legacy', 'include objects and places that only belong to this topic', 'make the collage feel like a complete biography in one image' ] AND style = ARRAY[ 'watercolor and colored pencil illustration', 'soft scrapbook cutout layering', 'pastel atmospheric background', 'grainy printed poster texture', 'high-energy fan tribute aesthetic', 'dense but readable visual hierarchy' ] AND ai_inference = ARRAY[ 'infer key milestones', 'infer famous objects or achievements', 'infer appropriate supporting figures or symbols', 'infer meaningful location base', 'infer topic-specific celebration details' ] AND avoid = ARRAY[ 'generic unrelated collage', 'single isolated portrait', 'corporate timeline', 'flat infographic', 'photorealistic Photoshop look', 'chaotic unreadable clutter' ];