I turned GPT Image 2 into a manga recommendation system. I used Death Note for this. Cam…

I turned GPT Image 2 into a manga recommendation system. I used Death Note for this. Came up with some interesting results. Function DrawMangaSmart(input title) Input Variable: [manga or book title] System Instruction: Generate a hyper-realistic, 1:1 product photo of an "Era-Aware Tankōbon Shrine Diorama". 1. Semantic Extraction (AI-Inferred): - AI infers: - Era (80s/90s/2000s/2010s+). - Demographic - Visual motifs - Figure: AI-inferred stylized Nendoroid-like figure of the main protagonist (not traced, but semantically accurate). - Pose: AI chooses pose based on genre (battle stance, shy posture, comedic flail, etc.). - Finish: Matte PVC. 2. Container (Spine Shrine Box): - Structure: U-shaped “bookshelf niche” with visible spines; floor is a book’s inside cover. 3. Background (AI-Inferred Panel Language): - Walls: B&W original manga-like panels AI composes to match era and genre (lineweight, screentone density). - Floor: Speech balloons, speedlines, and SFX cut-outs pop up. 4. Integration: - AI chooses a key trope (summon, confession, power-up) and turns a 2D panel of that trope into a 3D paper pop-out the figure interacts with. 5. Visual Syntax (AI-Adaptive): - Palette: B&W for panel world; full color for figure. - Era-aware tweaks: - Retro grit for 80s/90s. - Clean, high-contrast for modern series. - Lighting: Soft toy photography. - Label: "[Title] — [AI-Inferred Demographic Tag]" Return: ONE image, 1:1, Good-Smile-adjacent collectible realism. <instructions> Input: manga/anime title AI analyzes era, demographic, tropes. Find 4 MUCH lesser-known titles in the same micro-lane . Output: a 2x2 grid, each panel is DrawMangaSmart for a lesser-known recommendation. </instructions>