SDXL

Lucid Citrus Pony

#Furry

#Animal

#Fantasy

Lucid Citrus Pony - AI Model cover image
Subido en
14 nov 2024, 8:07
Utiliza
8.98k
Reseñas
Sin reseñas
Permisos
  • Allow image generation & sharing
  • Permita que los usuarios descarguen su modelo
  • Usos comerciales

Descripción

Current modelsKiwi (latest)The detailed furry step-siblingThe latest model in my Lucid Citrus series and a contribution for my animal-eared comrades: Kiwi. True, kiwi is technically not a citrus, but I am technically not a furry and wasn't planning on creating this originally. However, I wanted a challenging goal as a way to test new model merge methods. This is the result of multiple tertiary merges and meticulous layers of LoRA merges (to the point that I rewrote Kohya SS's LoRA merge script to use CUDA) -- all in effort to capture the furry elements of other semi-realistic models while excluding the human elements. My goal was a reliable semi-realistic model that would not need super heavy weights around "body fur" and "no humans" tags to avoid the human-skin/furry-face amalgamation issues while working on the weaknesses Clementine had when it came to generating realistic animals and furries.ClementineThe sporty jack-of-all-tradesThis is a semi-realistic model that is the result of a tertiary merge of multiple realistic Pony models and then a slight weighted merge with a few stylized 2.5d models. I was looking for a reliable bread-and-butter realistic model to sample LoRAs on and have been quite satisfied with it. It is really knowledgeable and recognizes a lot of tags and does not require as much post-step processes as much as other semi-reals do. The first six posted images are not flawless, but good examples of what it can do without adetailers, embeddings, or hi-res.YuzuThe unhinged daydreamer(Coming soon)Further details and information for each model can be found in the "About this version" section.Miscellaneous links for those who are interested:Modified sdxl_merge_lora.py to use CUDA instead of CPU for LoRA merges in Kohya SS. This is just a temporary home for it until I can work on it more to submit a PR (I need to add conditional handling to allow the option for both CUDA and CPU).My preferred LoRA block weight presets -- For those using the LoRA Block Weight plugin.

Comentarios1