Lisa Hunyuan Lora

example prompt:

• always starts with "video of a young woman with long blonde hair and bangs"

• long dark hair sometimes works too

• closed video short of a young woman with long blonde hair and bangs speaking

• video of a young woman with long blonde hair and bangs, She is dancing and smiling to the camera. The lighting is soft and natural, highlighting her features.

• video of a young woman with short black hair and bangs. she is applying lipstick in her right hand and applying it to her lips. The woman is wearing a black strapless top and her hair is styled in a bob with bangs framing her face. She has a serious expression on her face and is looking directly at the camera. The background is a solid purple color.

training info:

• this training done on images only, somehow hunyuan support that, it will become the first frame of the video, and then somehow works like I2V, based on your training image.

• training data is based on closed portrait, it causes the video always focus too closed to the face.

• I feels like the model is overtrained, you can tweak down the weight when the prompt adherence is bad

• training is done on RTX 4080 super with 16gb of vram and 32gb of system ram