> It is taken whole-cloth from an input, and reproduced as-is on an output. I respectfully disagree with this claim. It’s not as-is. It’s remarkably similar. T

> It is taken whole-cloth from an input, and reproduced as-is on an output. I respectfully disagree with this claim. It’s not as-is. It’s remarkably similar. That’s a big difference. > Stable diffusion is more like a paintshop artist who grabs bits and pieces of other art and melds them together, and less like a painter who creates from their imagination. I also disagree with this. The uncomfortable truth, imho, is that what Stable Diffusion does is FAR closer to what human artists do than we’d like to admit. Human artists are perfectly capable of reproducing copyright infringing images. They just generally choose not to for legal/moral reasons.