Base usage:
Sampler: Euler a
Steps: 24+
CFG Scale: 3.5~7.5、5.5
Hire.fix: R-ESRGAN 4x+ Anime6B、4x-UltraSharp、4x-AnimeSharp、、
Upscale factor：1.5、1.65
High score iteration steps: 10+
Denoising strength: 0.0~0.75
Clip skip: sdxl does not need to be set or any value.
Positive word:
Negative word:
very displeasing,displeasing,bad anatomy,artistic error,lowres,bad hands,signature,artist name,multiple views,artist sign,For better results, it is recommended to use ADetailer
https://huggingface.co/Bingsu/adetailer/blob/main/face_yolov9c.pt
https://huggingface.co/Bingsu/adetailer/blob/main/hand_yolov9c.pt
License: NoobAI License
Description
Known issue: characters from other series may be affected by wuthering waves character characteristics.
All wuthering waves images collected before October (45-2=43characters) have been trained.
Almost all wuthering waves characters can be generated, with character names in the form of danbooru.
In addition, three four-star male characters have been removed.
Currently, two characters are known to be unable to be generated correctly due to the limited number of images and the confusion of their features:
jianxin、youhu.
FAQ
Comments (4)
Hey, how updated is the Char-List on this Checkpoint? (Genshin, ZZZ, WuWa, etc.)
I love your LoRA's, did you just threw them all in?
Just trained all the wu wa characters. No other.
Not gonna complain about WuWa characteristics being intermingled with anything. Honestly I wanna see more variety in faces, bodies, and so on. If there was a way to have a checkpoint trained where it takes random anatomical attributes and generates a bunch of algorithms that randomly generate characters within a specified set of attributes that's be great. Say, "Pear shaped but with gaunt cheeks and lumpy skin."
Right now if you prompt that you're likely to get a beautiful super model that has the same euroasian face.
I don't quite understand.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.