Wan 2.2 / LoRA compatibility / Comfy native #39
Description
Eagerly awaiting the Wan 2.2 model release.
As y'all are working on it, will there be any better compatibility with other LoRAs? With the Wan 2.1 version, as soon as I introduce another LoRA to add action training, etc., the input face gets absolutely destroyed. Only running LoRA-free keeps the reference face integrity.
And any progress on a more "proper" native Comfy integration?
Great work on a novel concept!