It just tells me 'content not permitted'.

For context, I was attempting to put a cup of hot chocolate into the hands of an anime character.

Ever since OpenAI showed (but did not release) this type of multimodal output with 4o, I have been waiting for this to be available to the general public.

It seems like really combining visuals at the level of generation capability means language understanding is fully grounded in a richer world model.

I am hoping for a step up in real world common sense intelligence areas like those covered by SimpleBench. Although they are static images, so there might still be room for improvement ad far as physics understanding.

Also, if they can get it to the point of really accurate (probably larger models), this unlocks whole industries in terms of being able to do useful work.

I was really hoping that there would be more character consistency, given the fact they mention it in the blog. It also doesn't seem to reliably follow styles like "watercolor illustration" or "line and wash".