Whenever I access Hugginface’s Stable Diffusion models via ComfiUI, the models are lacking a safety filter; meaning they will generate harmful content, which does not happen if I use the code snipped to access Stable Diffusion models directly from Huggingface. In this case, harmful images will just be blacked out. Does anyone have an idea why this is the case? Why is the same model having a safety filter in one case but not the other?

I am trying to run the small tiny-stable-diffusion-pipe model from HuggingFace but it looks like it’s broken and won’t sample properly.