Can anyone help me with an image prompt / negative prompting issue? I am running image prompts that do not contain any faces or people, but no matter what I do they result in generations with faces
The behavior is so strange, I have been using prompts of architectural interiors with enzymezoo’s fork of huemin’s Jax 2.7- using prompt images with no faces, it still adds faces to most of the generations
It seems to do so in a very bizarre way, one interior it adds sleeping babies to, the other interior it seems to add men to from the geographic part of the world it assumes the image is from? (Notice the Buddha head) Please see the image prompts below
I have tried using negative weights of -1, -50, -100 etc for ‘face’ but it seems to always make faces
What can I do to make it stop creating face like forms?
I am using openAI + vitb16 +vitb32
Image prompt examples below + attaching images they generate above