Hoodwink character from game DOTA2
I trained and tested it with next models:
>Base model:
Anima_baseV10.safetensors
>VAE:
Qwen_image_VAE.safetensors
>Text_Encoder:
Anima_baseV10_txt.safetensors
Preferred weight: 0.75
Trigger word: xddwnk
It can generate both SFW and NSFW
It might sometimes generate a proper Hoodwink's crossbow, but I wouldn't bet on it.
Regarding version v2.0, please read the "About this version" section on the right side of the interface.
Description
This is LoRa version 2.0. I cannot explicitly state that it is BETTER or WORSE than the first version; however, testing has shown that it successfully mitigated the overfitting of v1.0 (either fully or partially). In my opinion, the model is generally more versatile. What does this versatility mean? The new version does not always attempt to render a hood, and swapping clothing is overall easier - it is more responsive to the prompt. I have not yet tested its compatibility with other (stylistic) LoRas.
Important changes: the trigger word is now ABSOLUTELY MANDATORY; for instance, write "Digital art of a xddwnk."
For her regular appearance: "She has brown fur, deep expressive blue eyes and dark tail with brown stripes and white tip. She has calm face expression. She has small a necklace in the shape of two green leaves. She is wearing green cowl, swamp green pants with brown armored plates and yellow sash."