Tried variety of prompts to attempt to narrow or trick Stable Diffusion into producing an exact replica of the original painting by Van Gogh titled “The Starry Night” — but all of them were in roughly the same style as one posted in the blog.
If anyone figures it out, please post as much info as possible: fork, image-set, settings, prompt, seed, etc.
Isn't this showing the artifacts of classifier-free guidance? I wonder if it would make more faithful reproductions if you turned it off?
Thats a good point. Let me try turning those off.
I don't get it, what input was used to generate these? The originals?
Their titles. Plus the training data, which would have many copies and variations of each painting.