import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("bytedance-research/HuMo", dtype=torch.bfloat16, device_map="cuda")
pipe.load_lora_weights("drozbay/HuMoveLora")
prompt = "-"
image = pipe(prompt).images[0]HuMoveLora
- Prompt
- -
Model description
A LoRA that tries to combine track controlled motion of Wan-Move with the human motion and speech sync features of HuMo.
Usage
Base model: Wan HuMo-17B Model
In ComfyUI: Likely needs WanExperiments to allow for I2V capabilities. Chain the HuMo node with the WanMove node. Recommended track strength: 1.5.
It is not recommended to use HuMo reference or start image along with the WanMove start image.
-
Version History
Version Notes v0.1 Proof of concept release. Further tuning should improve simultaneous speech and motion control capabilities.
Download model
Download them in the Files & versions tab.
- Downloads last month
- 4
Model tree for drozbay/HuMoveLora
Base model
bytedance-research/HuMo