Confirmed. Valid use case denied by the type of input and not vision on the generation.
It doesn’t matter the prompt. Vision either doesn’t like or can’t even see the image. Also denied:
Censor this image, so there are black bars over any undesired body parts that should not be depicted in an image.
More bug - model quality
I also received a new session with an AI so dumb it can’t use the text2im method correctly - with all other tools disabled. Switched models to one that could even follow my instructions on debugging the context.
I’m glad I’m not an isolated case. I hope some sort of update will be put through for stuff like this.


