Much like my facesitting LORA, this, like that one, was trained on the exact same dataset as the hunyuan version, using the exact same LORA trainer with the exact same settings and script.

However, on WAN, this LORA really, REALLY shines. On Hunyuan, a lot of times the effect wouldn't happen, because the spatial awareness wasn't as good. On this one, it REALLY works well lol.

YOU DO NOT NEED FULL 1.0 STRENGTH! WAN Lora seem to do well even at lower values! I recommend 0.6-0.8 for this one! Though 1.0 doesn't break it exactly either!