Creating compositions - the state of things today

[andupotorac]

I spent the entire day looking at what's out there, and it seems to me yours is the perfect tool, for unperfect tools. :-)

1. First off, I tried to use the NEW Composable Lora, the forked one since the original isn't maintained, and it gives an error (RuntimeError: The size of tensor a (80) must match the size of tensor b (78) at non-singleton dimension 2 a2569875/stable-diffusion-webui-composable-lora#15) when using it. So I reverted back to the old Composable Lora that at least doesn't generate an error.

2. The old Latent Couple just doesn't seem to take into account the areas given to it by using your tool. Or better said, it mixes them.

3. I even used Cutoff to prevent this from happening, but it just doesn't want to generate separate regions independently.

4. I'm also unsure about how the weights should work. Should their total be 1? Or can each be any number? Because it seems like the images get overdone when passing above 1.

This all seems problematic and I'm wondering what's your current setup. Are you using your tool / planning to release it for Regional Prompter too? That is at least maintained by the original dev and maybe it works.

My masks don't even overlap. This is all so frustrating.

These are the settings I'm using. I'm not even sure Composable Lora works ok, since it's the old version.