Heaven Burns Red
Style LoRA
(EN/日本語/中文)
|___________________________________________________________________________________________|
About this LoRA:
Trained with images of the popular mobile game >Heaven Burns Red<
___________
>Trained with >CouterfeitV3.0<
___________
>How to use it:
Personal checkpoint recommendation:
___________
-->Prompts: hbrwfs, blurry, blurry background, scenery, background
___________
-->LoRA strength: 0.8
-->Highres fix: Latent
-->Sampler: DPM++ 2M Karras
-->CFG scale: 7~11
________________________________________________________________________________________
Prompt example:
___________
Prompt:
masterpiece, best quality,hbrwfs, 1girl, solo, flower,skirt, indoors, scenery, night, window, bangs, black hair, black legwear, looking at viewer,blurry, blurry background,standing,uniform,night sky, horizon, stars,lonely, <lora:HbrStyleV.1:0.8>
___________
Negative:
1girl,badhandv5, EasyNegative, verybadimagenegative_v1.3,illustration, 3d, sepia, painting, cartoons, sketch, (worst quality:2), (low quality:2), (normal quality:2), lowres, bad anatomy, normal quality, ((monochrome)), ((grayscale)),
___________
*Important keywords are marked >BOLD<.
*重要なキーワードは >太字< で表記しています。
*重要的关键词被标记为 >粗体字<。
___________
________________________________________________________________________________________
Chinese and japanese instruction coming soon.
中国語、日本語は近日中に対応予定です。
中文和日文教学即将开始。
Description
FAQ
Comments (4)
Did not work for me, sadly looks like the style of the images comes from the base model and not so much from the LoRA
should not be the case. I've tried it out with various models and there definitely was a change in style. What model and VAE did you use?
@zeppelinofthe an anime model, similar to anylora and the VAE is the waifu-diffusion one. Is it more apparent with a specific VAE?
@victorc25744 Ok. I've tested AnyLora + waifu diffusion and with clearVAE. There is a significant difference so I would highly recommend to use clearVAE (https://civitai.com/models/22354?modelVersionId=26689). It is also the VAE that most of my LoRAs are trained on and is in general very good. Other changes you can do to get a clear difference: 1) generate horizontally, so 756x512 instead of 512x756. 2) make sure that you use the keywords "blurry, blurry background" in the Prompt. This should make your images much more Heaven Burns Red alike with its typical blurried scenery in the background.