
I just got Nano Banana v2 Pro and it's a big leap forward. You can now quite reliably edit images with text prompts! I have this test that stumped v1 where I ask the model '"Keep this image the same but fix the rope bridge so that it connects to the upper left treehouse platform on the edge instead of the bottom"' After several tries with v1 no success. It either moves the rope bridge around or does nothing. v2 nailed it first try. (before / after)