You can either spend millions on trying to beat the SOTA of Text-Image models or you can hire some designers and copywriters to make you a website and a paper, plus an AI ethicist to explain why you can give nobody access to the model. Nobody will be able to tell the difference.