I've been asked this a few times, so I thought it would be nice to gather the relevant information in one place and share some of what I've learned over the past months using AI tools. My basis is Automatic1111 which is a local StableDiffusion install. Using Dreambooth, I mixed Protogen 3.4, Realistic Vision, Cyberrealistic and DreamlikePhotoreal into "my" model, which I then finetuned on its own generations which I handpicked and fixed the most glaring issues. I do not intend to release this model which I simply call JV1, not because I consider it my own or for some elitist reason, but because it is tailored to my needs and will probably not work too well for others. I'd rather give people the knowledge of creating their own models which suit them. Once I was happy with JV1, I generated some faces and used the best ones to create an embedding for my OC, which I use to generate consistent faces. I also created an embedding for midsections, to avoid some issue you can run into, e.g.