V2.0:
The new model's dataset has been optimized, significantly reducing overfitting and improving the accuracy of prompt-based generation. However, the reproduction of clothing details may be slightly less precise than in the previous version.
Suggested weight:0.6-0.8
Trigger words:
Character
Sengoku Kaname,blonde short hair,brown eyes,1boy,male focus,solo,
I like to exaggerate a character's traits by adding
bara,mature,muscular,
Outfit
Outfit1(brown jacket)
brown jacket, black t-shirt, white pants, black boots,
Outfit2(white jacket)
white jacket, black shirt, belt, brown pants,
Outfit3(sweater)
blue sweater,grey pants,
TIPS:
Most training data comes from animations, and the result can be enhanced by adding artist tags.
Description
The dataset of the new model version has become more streamlined, reducing overfitting and enabling better generation based on prompts. However, the details of clothing may not be as accurately reproduced as in the previous version.