I investigated the question whether a generative AI model can be trained to generate images for book covers, with no other input besides the synopsis of the book. Here’s the result: An essentialist model, based on Stable Diffusion 1.5. Unlike the SD base model, this model creates aesthetically pleasing cover art with space for adding text.
The custom model is on the right, the standard model it was based on is on the left to highlight the difference. Parameters are the same; input is the first paragraph of a book synopsis.