Closed drago87 closed 11 months ago
Interesting. What's strange is if you take the latent out of the Empty Latent by Ratio node and just decode it, it's the correct resolution (512x768), meaning that the adj_width
and adj_height
are working for creation of the latent image but not for the output width/height. The floor division by 8 is necessary for the latent, but not the integer we want to return. To make sure the output width/height is accurate to the image generated and not the input, I'll fix the return for those integers to effectively be adj_width * 8
and adj_height * 8
, which should match Thanks for finding this!
When reading the int from the node i get this if i run the latent to a ksampler and make a image i get the right size
and i think it is
adj_width = width // 8
row 550adj_height = height // 8
row 551that somehow breaks it. or you should return the
width,height
instead of the adjusted values (if there is not something i'm missing and it is not for the image size and something else. if so something that indicates that would be nice)