This is a model trained on forest. Trained to add detail to the normal nature background. May need to increase it starting at 0.6 for desired results. This model was trained on NAI anime-full-pruned and generated in Based68, and cutesweetmix. Made with the Kohya tool for AI Training through google colab. Any suggestions to improve or recommended things for the data set would be appreciated.
Description
This is a tough one since most checkpoints have an idea of what a forest is So I wasn't satisfied until several test using the same checkpoint several times same prompt comparing with and without the lora. This lora from what I've seen hides the normal random path and road if used in the negative tags or you can prompt for them as you wish a problem I see a lot on the raw generations. It doesn't shade out the trees for long dark lines and includes branches and leaves even in night or foggy images still showing appropriate detail given the situation. This was trained on real images, game backgrounds (when given knowledge of libraries of game resources I used it for nature shots...), paintings, and pixel art (a mix consisting of about 60 images). It has good detail in general with fog and moss adding some neat control for images.