This is an awesome model. Since this was trained as a style, I think it's better suited for img2img instead of txt2img. The initial image was generated with Juggernaut final, then run through img2img with Analog Diffusion, the "analog style" tag applied, and a denoise of 0.4, then a denoise of 0.75
This is a super fun model! I even did decent photos of Lewis Carroll. Other models made his awful fat guy or Edgar Allan Poe. Even using Charles Lutwidge Dodgson didn't help. There aren't enough images of online. Analog couldn't fix that, but amusingly enough, it generated a guy who looks exactly like someone hollywood would cast to play him in a movie. 🤣. The best it could do and pretty darned funny. Also, the image quality was pretty good too.
I love look of this model, also one of the most realistic ones I used. It can do semi stylized stuff as well.
Shorter prompts tents to work better (at least for me)
Its really good at details but during upscaling to high resolutions it tents to sometimes do artefacts.
A go to for retro look!