toyxyzさんのイラストまとめ

ControlNet Semantic Segmentation test. #stablediffusion #AIイラスト #seg2image

ControlNet Semantic Segmentation test. #stablediffusion #AIイラスト #seg2image

ControlNet Semantic Segmentation test. #stablediffusion #AIイラスト #seg2image

Through this method, mixing of people and objects can be prevented to some extent.

But there are limits. In the upper image where the prompt is "1girl", the 2nd girl became a strange lump, but in the lower "2girl" it was generated normally.

By arranging #0066c8, which is an object (such as a car), #04c803, which is a tree, and #96053d, which is a person, it is possible to designate a relatively more accurate location than the usual method.

ControlNet Depth2animation test #stablediffusion #AIイラスト #depth2image

ControlNet Segmentation animation test #stablediffusion #AIイラスト #seg2image

ControlNet Semantic Segmentation test #stablediffusion #AIイラスト #seg2image