Slight tangent, but have you considered using these new generative models like Dall E 2[1]? So far my experience has been that they generate great oil paintings. Maybe there could also be an automated pipeline that generates oil painting covers using book titles or first sentences.
Leaving aside the aesthetic considerations, the images produced from OpenAI software aren’t public domain, and so we can’t use them. See https://openai.com/api/policies/sharing-publication/ for a list of restrictions OpenAI places on the use of their generated images.
[1] https://openai.com/dall-e-2/