I could be wrong but it seems to me that denoise max is much more detailed than before.
I need to do more testing on this.
The only concern i’ve is that i’m able to spot something like a repeating cross “X” pattern in grass and other surfaces.
I don’t know if its the architecture or not enough grass is in the training pool.
I know something like the “X” pattern from other diffusion models, like images generated with chatgpt.
