That should make it able to use web searches and other images in order to understand and reason its way into better pictures.
Like with Nano Banana Pro, you could create diagrams from notes, make infographics from text, and «generate data visualizations,» only a whole lot faster — and cheaper.
It should also be better at subject consistency, more precise in following instructions and can create high fidelity images at up to 4K resolutions.
It instantly leapt to the top of LMArena’s (now just arena.ai) text-to-image leaderboard.
Read more: Google’s presentation. Writeups on 9to5Google, Ars Technica, and TechCrunch.