This hypernetwork transforms landscapes into "smoother" variants and more "dreamy" at positive strength (1.0), while adding roughness, crispiness, and more dirty detail at negative strength (-1.0)! Example images show both so you can compare. Trained on deliberate_v5 model, but tested and works on other models just as well.

It also seems to have a subtle effect for non-landscape prompts, including characters. However, its training data consisted solely of landscapes (3k+ of them), so keep that in mind.

Feel free to experiment with strength/weights, including negative ones.

I'm trying to explore hidden potential of hypernetworks.

Support me on Ko-fi