
Handsome Lad | SD 3.5L | Robbie's Original
-
For SD 3.5L Q8 and similar | use your standard, whether Euler 15 and 4, or other settings.
-
For SD 3.5L Turbo | you can start getting results at 5 steps and CFG 1 with Euler/A, then increase.
-
For full effect you can go up 0.8 to 1.X (But up to you.)
-
For more freedom and maybe art styles, lower it down around 0.5 ~ 0.8
Objective here is not exactly the same as Handsome Lad for Flux. It does have different facial hair and hairstyle (AND THEY WORK HERE AS WELL YAY,) Still you can get some interesting results.
Works for portraits, fantasy, etc. Despite Flux being good for realism, I really like Fantasy and stuff on SD 3.5L.
It should work on fine-tuned checkpoints. Some are amazing, others too biased and there's nothing I can do.
Prompt for those training image samples were simple: "Image of a handsome man, professional portrait, highres, high-definition, colorful, smooth, clear, clean"
Prompt for those samples, read the images.
Dataset is NOT the same used for Flux neither SD3(?)
Disclaimer: I cannot control deformations which are typical from these latest SD.
描述:
Trigger words are optional
训练词语: handsome lad,man
名称: handsomeladstypeCSD35L.safetensors
大小 (KB): 263415
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success