Problems
A little hesitant to be flexible with prompts: if anyone has any suggestions on how to fix this, please let me know.
Hats appear too often: A good portion of the images used to train this model were ones with Elezabeth II wearing hats. If you do not like the hat, then use negative prompts like "hat" or "crown" or even "tiara".
Suggested Settings
Sampling Steps: Anything above 20
Size: This model was trained with 768x768 images. Anywhere from 512-768 will work.
CFG Scale: I found that 7 works significantly well. Whenever I would lower this from 7 it would make her look significantly younger. If anyone knows why this happens just let me know.
CLIP SKIP: Clip skip 2 is HIGHLY recommended to anyone who wants an accurate look.
Sampling Method: I use DPM++ 2M Karras, however, any sampler could work.
Other Information
Keyword is T8EQU33N
If anyone finds any problems, please notify me in the comments.
描述:
训练词语: T8EQU33N
名称: Elizabeth2_LORA.safetensors
大小 (KB): 147570
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
