Been messing around with Nano Banana (that Gemini-series image gen + editing tool), and the biggest takeaway is that it actually reads you before generating—not just slapping a bunch of keywords together. It can do 4K, any aspect ratio you want, 16:9, 9:16, 21:9, all supported, and you can feed it a dozen reference images plus text all at once.
As a veteran designer, what I appreciate most is this: writing prompts is like giving a brief to a photographer. Stop mindlessly listing style keywords. Start with a clear verb to tell it the main action. For text-to-image, I usually follow the skeleton of “subject + action + scene + composition + style.” For example: “A model in a custom brown dress, confident stance, deep cherry red studio backdrop, medium close-up centered composition, high-fashion editorial vibe, medium format film grain.” That one sentence gives you way more control. Using reference images is better for keeping character consistency or sticking a product into a new scene. Structured descriptions—that’s a habit I picked up from client work.