A LoRA for Hunyuan, trained on ~250 images from high-end lingerie brands. The dataset had faces removed, making this LoRA ideal for LORA stacking or use with character LoRAs to maintain consistent styling.
Description
Initial release.
FAQ
Comments (4)
Muchly needed LORA. Any chance you could provide the captioning information or a list of unique phrases?
Glad you like it! Would love to see some of your gens. Data uploaded.
The training captions are helpful, but would it be too much to ask for the images as well?
So, after playing around with v1.0 (and comparing it to generations from just the base model), I've found that this LoRA adds a little "frisk" to the person's movements and heightens the details on the clothing.
I've found using it at strength 0.4 (you don't even need the trigger word) and pairing it with a lingerie wildcard like https://civitai.com/models/456442/dynamic-wildcard-lingerie-unleashed or https://civitai.com/models/613244/luxury-lingerie-wildcard gives good results.