Consistent Face (Wan Video 14B t2v)

• Trained on 3D video to enforce a consistent face across T2V or I2V

• In 20 or so runs the face was consistent, however that does not mean that it would be that for 100% of the generations. Seeing as the cost of those runs was 10k buzz consider reading the following before posting another what does this lora do comment.

I don't know how much more clearly I can explain I can not predict the full effects of any Wan LORA I train when at best I get a few epoch images, and at worst I get no information at all.

For my local trainings I can monitor gradient, loss, I can evaluate 90/10 or 80/20, I can and do generate thousands of images to test. I DO NOT HAVE THAT LUXURY CURRENTLY WITH 14B