Would it be possible?
I've looked at the training parameters for Wan2.2, and it doesn't support multi-image control. This means that everything depends on the model's own capabilities, which seems quite difficult to control.
· Sign up or log in to comment
