Another new model, another new style: here's a quick first play around with OpenAI's GPT-image-1 on CivitAI.

Started with a simple prompt of guys in a single style of raingear.

Then tried a few hikers in classic nylon cagoules and quickly discovered how responsive and reliable it is with complex prompts. It's remarkable how good it is compared to the way things were a couple of years ago, when you had to generate a dozen images just to pick out one decent one worth keeping. I'm posting every one of my quick test gens here, no selective editing. Some of them aren't perfect, but none of them are really obvious fails.

So I had to finish with a few back on the imaginary yacht.