I’m trying to understand why in Automatic1111, i can use a very low denoise factor (0.2-0.3) and the images do not come noisy or badly rendered, but in diffusers, when i do TXT2IMG + UPSCALE + IMG2IMG (hi-res), no matter what upscaler I use, and i’m using the same as automatic, if i pass a value lower than 0.4 to the strength attribute of the StableDiffusionImg2ImgPipeline, it begins to fill the image with artifacts.
Any ideas, has anybody found this issue?