I don’t know if anyone has the same problem: when I use the controlnet inpainting model via diffusers StableDiffusionXLControlNetInpaintPipeline, the result didn’t follow the prompt, but the same model works well on the webui.This is only happens with XL version(no matter the mask area of control image is set to 0 or -1, the original controlnet uses -1, so does the v1.x diffusers piepeline, xl diffusers piepeline change it to 0.Would that be more reasonable? Why change?)
Besides, I found the results with/without cotrolnet are weird，can’t explain why it happened.
1.base model: dreamshaper-xl-1-0 (I downloaded the model from huggingface for diffusers and civitai for webui, I guess there’s no difference.)
2. controlnet trained with diffusers
3. prompt: a blue suv is driving down the road
The result of diffusers is completely unexplained: the trained controlnet makes the result worse.
