Wow! It feels like I grabbed my 6D Mark II camera and was actually on location, guiding the model on how to pose to make it look beautiful.
The image generation control from the GPT-4o model is incredibly precise. It brought to life everything I asked for — from the spatial orientation relative to the sun, to the direction of the wind that lifts her hair.
Of course, there are some flaws — like imperfect facial consistency or the size of the watch on her wrist. But maybe I just haven’t figured out all the tricks yet.
Overall, for single image generation, this is the most powerful model I've seen so far. For generating a whole series — still needs some learning.