Gemini can generate normal maps for your 2D artwork, so it’s a cheap and easy way to generate 3D effects with AI.
This was for an overly ambitious project that I probably won’t finish, so now it lives on as a blog post.
I was using Google Gemini to generate art for a trading card game mobile app. I’m always thinking about how to get more “bang for my buck” out of generative AI. It’s great at generating 2D art Yes, I know there are people who absolutely hate AI-generated art, but I mean it in the sense that I definitely am not able to draw this. , which is what led me to the trading card game concept in the first place - no complex 3D models needed, but we can put the 2D assets in 3D scenes with shaders and effects, etc.
After making the 3D viewer for the cards, I was experimenting with typical trading card effects, like holographic foil. Then I thought: what would really make the cards come to life? What if they had depth in 3D?
I don’t have the time, talent or inclination to make accurate 3D renders of these hyper-realistic portraits. So I attached the source artwork in my Google Gemini prompt, asking for a normal map, and sure enough, out it comes My metric for how accurate these normal maps are is whether they look cool to me or not. In that sense, I am essentially the first vibe normal mapper. . It was just a case of asking Claude Code to apply normal maps in the 3D scene after that.