Tipo
LORA
Tempo de Publicação
2024-11-16
Modelo básico
Flux.1 D
Palavra de ativação
Bootsy Collins
Copiar
Introdução à Versão
The dataset consisted of 16 images, mostly from the 1970s and 1980s. Five of the training images were up-scaled with SUPIR and manually corrected. One image was up-scaled with 4xFaceUpSharpDAT. The images were simply captioned "Bootsy Collins".
As the focus was on his Bootsy himself, and the training data under-represented and scarce, the LoRA may fail to capture his gold tooth, costumes or star-shaped bass accurately. Unless instructed otherwise, the LoRA does have a tendency to make him a bit plumper/older looking, as likely the base model mostly saw older pictures of him.
