Matt Wolfe tests Gemini's new image generation and editing tools. It creates a bilingual cafe menu with legible small text and correct prices, then uses doodles as editing instructions to transform sketches into photorealistic 3D renders with proper lighting and materials.
Well, Google did it again. Nano Banana too. So I gave it the prompt. Design a bilingual cafe menu with English and Japanese in a minimalist style with perfectly spelled ingredient lists and prices and illustrated icons made from clay. This was the output here. And we can see we've got English and we've got Japanese. A lot of times when you have really small text in these images with these various image models, it doesn't actually read well. It's usually just gibberish. But on this, even the teeny tiny text below that describes the ingredients of this is all very coherent. You can also doodle and modify photos directly inside of Gemini. So here's an image of a person wearing an Apple Vision Pro. Let's give it some horns like this and let's make the eyes much more visible, like so. And I'll give it the prompt, use my doodles as instructions, and redesign this project with the features I pointed to. Make it look like a polished marketing image and perfect. Exactly what I was hoping for. Transform this sketch into a photorealistic 3D render of a consumer tech gadget including materials, lighting and a realistic product hero shot. And just like that, it took this sketch that I totally drew myself and turned it into a pretty good looking realistic image. There's probably some better image generation models, but as far as image editing goes, this one pretty much crushes everything right now.
AI News: Dots, GPT-6.1 Sol, Sonnet 5.5, Gemini 4, and everything you need to know
