Is this a distillation of Nano Banana Pro?

Open weight? How many parameters?

It's not working very well at all. I started with a picture of a girl sitting at a cafe table and asked it to zoom in, and it enlarged her head to the size of a balloon.

does it still break images with transparent pixels?

Still has context leaking into the text/random signs in the image, made worse by generating filler with an internal LLM

Just what we need, more sloperators thinking they are being creative and making art by prompting.

I would be happy to never see any more AI slop.

Can we now edit the images it spits out? All prior tests in trying to edit AI images has failed miserably and laughably

Is it just me, or is Nano banana not working in Gemini currently?

The prompt “Generate an image of an orange man riding a bicycle.” no longer generates Trump riding a bicycle. I’m guessing they fixed that.

But the prompt "can you depict a cartoonish orange man with a pooh bear in political cartoon style?” correctly generates Trump.[1] So there’s that.

[1]: https://imgur.com/a/Gvm5Zje

It's working pretty well for generating an xkcd comic for your HN profile: https://hn-wrapped.kadoa.com/

Previous nano banana frequently made speech attribution errors, the new one seems a lot more consistent.