My third attempt at making a style that is a composite of characters from Aobuta/Bunny Girl Senpai. This one has much more promise than the last two. The first part of this version was trained from 309 auto-captioned images of Mai, Rio and Shouko. The second part of this version is trained on 247 auto-captioned anime images of Tomoe, Nodoka and Kaede. All images were selected from anime image datasets for clarity of facial expressions.
There will probably be a third part which combines all of the main girls.
Down the line I might want to make a Lora that is also trained from images from the other anime that Satomi Tamura did character designs for, The Pet Girl of Sakurasou, which have a similar style.
Wow, Pony does so much better with this than XL. V3 was trained from almost the same image set as V2, but done in Pony instead of XL, and it works so much better!
描述:
This is a version of the Aobuta style that was trained on Tomoe, Nodoka and Kaede. It was trained from 247 auto-captioned images from the anime selected for facial clarity.
训练词语:
名称: Tomoe_Nodoka_Kaede.safetensors
大小 (KB): 223109
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
