This is the Huyuan video version of my previous Pony XL version, which uses the same dataset resized to 512px but JoyCaption-Alpha-two for captioning.

This Lora produces decent results as long as you follow a prompt like this example:

Life-like video of an Asian man with long, straight black hair tied in a high ponytail, wearing a traditional teal silk robe adorned with white birds motifs. He has a fair complexion and a serene expression. The man's attire includes a white undergarment and a brown leather belt with intricate designs, indicative of historical or cultural significance.

This is the "2.0" version already, I tried to train the first one using just tags, it worked but the outputs were too hard to control.