pony-diffusion-g5-lora - LoCon saves the Unity Crystals

pony-diffusion-g5-lora is a latent text-to-image diffusion model that has been conditioned on high quality pony images through fine-tuning using LoCon from LyCORIS's implementation.

Finetuned for MLP G5 characters, based on AstraliteHeart/pony-diffusion. No furries, no NSFWs, just pure pones

This model is LoRA-only. See Files tab for downloads.

!! IMPORTANT: Add tag "pony" in prompt can DRASTICALLY improve image quality

(ignore his red hair plz it's a teeny tiny training error)

Dataset criteria

All training images are from Derpibooru using the search criteria below

"{character_name}, safe, solo, score.gte:150, -webm, -animated, -oc, -g4 || {character_name}, suggestive, solo, score.gte:150, -webm, -animated, -oc, -g4 "

where {character_name} will be replaced by one of the following: sunny starscout, izzy moonbow, pipp petals, zipp storm, hitch trailblazer, misty

How to use LoCon in A1111's sd-WebUI

Since LyCORIS's implementation isn't support by WebUI officially, you have to load it through extensions like a1111-sd-webui-lycoris or sd-webui-additional-networks. Also, you have to use AstraliteHeart/pony-diffusion because these LoRAs are based on it.

Thanks

AstraliteHeart/pony-diffusion, for providing a solid start-point to train on

This project would not have been possible without the incredible work by the CompVis Researchers.

With special thanks to Waifu-Diffusion for providing finetuning expertise and Novel AI for providing necessary compute.

---

license: bigscience-bloom-rail-1.0