I work in the AI field, but not on image generation.
I don't think it would be technically hard to build a model with current technology which can generate logos with the attributes which you mentioned. You could simply fine-tune a Dalle-E style model specifically on a smaller dataset of logos. This would just take a small dedicated team of domain experts to work on the problem.
I don't think it would be technically hard to build a model with current technology which can generate logos with the attributes which you mentioned. You could simply fine-tune a Dalle-E style model specifically on a smaller dataset of logos. This would just take a small dedicated team of domain experts to work on the problem.