Fursuit lora, my first Klein 9b loraTrigger word: f3rsuitKey: Download both models, highly reccomended NOTE: ALL SAMPLE IMAGES ARE CREATED USING UNDISTILLED FORESKIN MODEL + BASE TO TURBO LORA That aside, I will be uploading multiple of the loras, it turned out a group of them actually work very well depending on what you are trying to achieve. -For pure T2I output, lower epochs (single lora mounted) at 1.0 value is generally the most stable, but the quality of the fursuits are much lower than those of the higher epochs, Solution? Use the base image edit workflow, mount a high value epoch, use between 0.75-1.0 (optional, mount an additional LOW epoch at 0.2-0.25 in conjunction with the HIGH epoch at 0.75), and simply mask the suit head and I2I it, of course including the trigger word in your prompt, -For best quality-of-fursuit output, use either a HIGH lora at 1.0 mounted alone OR combine it, like above, with a LOW lora at 0.25, so then you mount the HIGH at 0.75, gives you a balance between quality and stability LIMITATIONS:Due to overfitting with the higher epochs, human faces will become massively distorted unless you lower the value of the strength, OR, you can simple I2I the failed faces, unmount lora or drop it to 0.2, and then just replace their faces with a new face. Overall, try to avoid masses of humans inside of the image unless you're purely using the lower epochs, but that aside, you can achieve some pretty awesome results and I myself do enjoy the lora and plan on making more klein lora.

Files with SHA256: 5dc5b42cb71e82bffa792096b8c74d1229fc446873508e4befb3f706624652f4