OpenAI just dropped Image 1.5, and I tested it properly—realism, text accuracy, speed, and (most importantly) control. I walk through the new “create image” interface with embedded styles, show how fast it generates in real time, and then push it with transformations, mixed-media edits, and instruction-heavy prompts. The big surprise for me is how well it preserves details during precise edits—this is the exact area where GPT Image used to let itself down, and where Nano Banana Pro has been strong for a while. Then I put Image 1.5 head-to-head with Google Nano Banana Pro using the same prompts: overflowing wine (a classic dataset bias test), dense typography, grids for consistency, and a bunch of community comparisons. My takeaway is that both models are now extremely capable, and the differences are genuinely nuanced. If you care about clean typography, intricate graphics, and usable design output, Image 1.5 is marginally better. If you care more about stylization, mood, and cinematic vibe, Nano Banana Pro still has the edge—and I still think Midjourney deserves a mention for pure aesthetics, even if its editing isn’t as powerful.