This is NOT my model. Made by waikitty on Civitai, this is just a reprint.
If you're the creator of this model and want this reprint taken down, leave a comment below and I will do so.
Prompting:
Quality Tags: masterpiece, best quality, newest, very awa, absurdres, ultra detailed,
Negative Tags: worst quality, bad quality, normal quality, lowres, old, oldest, worst aesthetic, bad anatomy, bad hands, extra digits, fewer digits, bad feet, bad eyes, distorted, disfigured, error, jpeg artifacts, sketch, text, watermark, signature
Your positive prompt should be structured like this from front to back:
quality & aesthetic tags, artist (s), 1girl/1boy, character, series, general tags
Positive prompt example: "masterpiece, best quality, newest, very awa, absurdres, ultra detailed, ciloranko, ask \(askzy\), 1girl, ganyu \(genshin impact\), genshin impact, solo, blue hair, ahoge, purple eyes, smile, bodystocking, gold trim, cowbell, goat horns, looking at viewer, standing, contrapposto, hand on own hip, feet out of frame, dark background, light particles"
Use Danbooru tags to word your prompts -> Danbooru Tag Groups Wiki
When adding artist tags, do not use "by" or "artist:", simply put the artist name
Do not use underscores anywhere in your prompt unless they are for a LoRA trigger word
Settings:
No VAE
Use SDXL resolutions: 1024x1024, 896x1152, 832x1216, 768x1344, 640x1536
Euler a
25-35 steps
2-6 CFG. Anything higher than 6 tends to create oversaturated images. This can be partially resolved by adding "saturated" and "high contrast" to the negative prompt
Clip Skip 2
For upscaling, use 4x Animesharp with 0.4 denoising strength