This is a high-noise LORA model trained on the wan Video 2.2 I2V-A14B base model. It can be applied to both high-noise and low-noise pipelines simultaneously, with the model weight set to 1.0.
Its function is to realize this specific dance movement—you can use it solely for entertainment purposes or incorporate it into your film and television productions.
The model was trained for 6 hours using an RTX 4090 24G graphics card, and this version corresponds to the 2700th training step. A total of 9 vertical-resolution materials were used in the training process—thus, it performs best when used with vertical resolution.
In my tests, the model shows good compatibility with clothing styles like short-sleeved tops and shorts.
Use the following text as its trigger phrase:
(HUPODBC，画面中女性双手轻放于颈部后方，头部与身体随节奏做出动态变化，呈现出舞蹈或展示类动作，整体姿态自然且富有韵律感。)
The video displayed on the page was generated using this LORA with the quantized Q8 model, running for 8 steps. If you use a better base model and higher step counts, you can definitely create even higher-quality videos—feel free to experiment with different configurations!
Local usage requires high computer configuration. If you want to have a try, you can click the link below—I have turned this LORA into an AI application and hosted it there. By registering through my link, you will get 1,000 points, and you will also receive 100 points every day when you log in. With these points, you can generate plenty of images and videos for free!
https://www.runninghub.ai/ai-detail/1960327801326940162/?inviteCode=rh-v1158