LORA by Royal Candy in Smile Precure, learning through official and fan art.
Trigger: 'loyalcandy'
Head
(loyalcandy:1.2),blue eyes,double bun,blonde hair,hair bun,short hairclothes
magical girl,pink choker,choker,jewelry,wings,brooch,white wings,bloomers,white footwear,white boots,wrist cuffs,tiara,earrings,white bloomers,puffy sleeves,white dress,yellow dress,skirtWeight: Between 0.8 / 1.0
Description
First version. Training in 26 images.
FAQ
Comments (6)
Hello. What checkpoints were used in the sample pictures? I tried the resources you suggest and a few others, but they do not result in such pretty pictures ;-(
Thank you for reporting the defect. The main checkpoint used in the sample is Hassaku XL V1.3. Just to be sure, we also checked the operation with another Pony checkpoint, ‘AutismMix SDXL’, and found no particular image corruption. As a side note, all sample images use the upscaler by Hires. The upscaler is ‘R-ESRGAN 4x+ Anime6B’ and the denoise is 0.5, which is 2x magnification.
If this still does not produce a good clean image, could you please tell us what environment you are using and what prompts and settings you are using?
@shadow_mochi I use the civitai image generation with the option "remix" thus using the settings provided by you. With Hassaku XL V1.3 and Workflow "Text-to-image Hi-res fix" with 0.5 denoise, I get similar pictures but they still do not reach your quality. It might be the upscaler that is different.
@Fluffia_Ai
The images were generated using CIVITAI's generation function, and the quality of the images was almost as good as the sample images, among other things.
@shadow_mochi Thank you. :)