stabilityai

/

stable-diffusion-xl-base-1.0

like 5.3k

I'm pretty sure the problem isn't the model but the users... The model only does whatever corresponds to the prompt. It should be rare enough for those kind of images to appear, as the model was trained with tons of images of different kinds it still might happen to make photos that might be problematic, but this is usually not an issue about the model. The model is... well... a base, it works perfectly for what it was supposed to be used for, but you shouldn't use it for a finished product, you should instead train it once again with other data, or, in case you want to avoid specific content, use a filter to remove any problematic output.

I hope I do not sound mean, it's just that at some point It's a bit discouraging to see people complaining about "a product doing what's supposed to do", if you ask it for something and it does it, it's not because the product is bad, is it? Still, I understand the problem so that's why I feel like you should just use a different model that was fine tuned to avoid that type of content, or to use a filter that removes that type of content of being displayed. The model itself does not have a filter to differentiate what content is problematic or not, it just makes up images that corresponds to the prompt after all...