Some highlights of Google's image generation model, gemini-3-pro-image aka nano-banana-pro. These are *not* the overall best examples of the model, but rather the best that I have prompted. Click the images for full size, and mouse over for the prompts.

It's really good at alliterations

Write an alliteration that is at least 100 words in length. Avoid using words where the first letter breaks the pattern. Then, make an image depicting the scene described by the alliteration, and put the alliteration text somewhere in the image that makes sense semantically (such that the text is part of the scene)

Why vibe code with Cursor when you can vibe code with Nano Banana Pro?

make a screenshot (DIRECT SCREENSHOT, NOT a photo of the computer) of C code being edited in vim, with syntax highlighting now implement the logic (by producing an updated image) instead of just showing the updated code, show a diff inline of old vs new code add a bubblesort impl

Useful for visualizing new/joke ideas quickly

image of inside of a plane, where, on on the left side of the aisle, about half of the seats are taken; it looks like normal. on the right side of the aisle, there is the same number of people, but only half the seats, so all the seats are occupied and there is twice as much space between each row of seats. make it a technical diagram style, such as a proposition of a novel comfort-maximizing seat technology where seats (demonstrated on right side) can be slid along a track so as to stow the unused ones in the back; dynamic seating
a long string of paperclips bent to spell out 'Banana Pro is a very impressive model'

It understands mirrors

Photo of a nicely designed bedroom. A newspaper rests on the table below a mirror. The text on the newspaper is 'Old man yells at cloud.' The text is reversed in the mirror.