A Lora trained on NAI with 36 images.
weights 0.7 - 1
some negative embeddings such as fastNegative can affect them.

think the kind transparent-ness in some of the sample images is due to fastnegative