A LORA of Alba B. belatedly uploaded, by request. Trained using the Kohya_ss gui on about 30 images on vanilla SD 1.5. This was my 4th attempt and it works fairly well with various realistic and Anime based models. Can be a bit frustrating at high prompt weights and I would suggest Clip skip 2 and using DPM++ 2M Karras as the sampler. Prompt info should be attached to the images.
There are no tags just use "<lora:Albapt2:1>" in your prompt somewhere. Prompt strength should be anywhere from 0.5 to 0.7 though some models will work with the full weight. It is a bit over-trained I think, though it is flexible enough to change hair, eye, and skin color and to add expressions.
描述:
训练词语:
名称: Albapt2.safetensors
大小 (KB): 147573
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
