1/8/25: Added WAI-NSFW-illustrious-SDXL version with new, improved data set, and a list of tags that were in data set. See version notes for more details.
Trained on 21 images from the animated film 'Hunchback of Notre Dame,' this is the Illustrious XL version of Esmeralda.
She works best around 1 strength. If you have trouble with the skin tone, I found adding "dark skin" into my prompt helped.
Like my work? Please consider leaving a tip to help fund future LoRAs or commission a custom LoRA for yourself. Higher quality versions of my LoRAs are also available for a monthly donation: https://ko-fi.com/artfulgenie.
Description
As with Jasmine, this one has a brand new and improved data set and was trained on WAI-NSFW-illustrious-SDXL v8.0 (https://civitai.com/models/827184/wai-nsfw-illustrious-sdxl). Because the WAI-NSFW model was trained at 1360x1360, I trained this Esmeralda checkpoint at the same size and it should be able to make some large images with pretty clean lines.
FAQ
Comments (3)
Is this Lora or Lycoris? Promp shows lora and description lycoris
So LyCORIS is a version of LoRA with a lot of variations under it. This is spicifically a LyCORIS LoCon here is a little write up for you https://stable-diffusion-art.com/lycoris/
LoCon allows for some of the unet to be changed when it is patched to the model, and it offers more space in the model for knowledge of what you trained, which hopefully means a better representation.
They just don't offer the option for me to say it's LoCon. Also in kohya_ss you can train these types in the options but this lora was actually directly trained to the checkpoint and then using one of the utilities in the back of kohya_ss we subtract off the original model to create this lora. (Esmeralda WAS finetune) - (WAS checkpoint) = (left over difference that is the training data in LoCon form).
😍 👍🏻
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.