New for v2!
Trained on SD1.5 base
Larger dataset.
All-new dataset (generated from a different prompt). Full set at https://huggingface.co/datasets/ppbrown/faeryqueen
Important notes
For some odd training reason, putting "red eyes" in the negative, often normalizes things. Your choice of model determines whether it is needed or not.
Some models do best when you drop to cfg=4. Others do not.
This is a fairly simple LoRA. Sort of like an embedding, really. Pushes a character towards the style of a "Faery Queen". Elegant ethereal female with wings.
Not a particularly amazing LoRA. I made it primarily to learn how to make them. If you would like to know how I made it, check out my article!
https://civarchive.com/articles/4789
Interestingly, I split it into two "concepts", and now it requires a trigger word "faery", whereas before, it didnt really need it
Description
I tried doing a buncha different things this time.
one of which is training on SD1.5 BASE model!
Plus, drastically changed my prompt for source images. Lemme know what you think.