You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The following install guide is by default for linux. For windows users, things become a little complicate. I prepare a standalone install guide for it Windows Install
How to Install on linux
This repo contains the required nodes and UI settings to run LHM on custom videos and custom input human images. Let's see how it going~
# If you do not install comfyui before
git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
Clone && Install LHM
git clone --branch feat/comfyui https://github.com/aigc3d/LHM
# link the nodes under the custom_nodes(using absolute path!)# If you install ComfyUI in /home/workspace, then /path/to equals /home/workspace.
ln -s /path/to/ComfyUI/LHM/ComfyUI-LHM /path/to/ComfyUI/custom_nodes/ComfyUI-LHM
ln -s /path/to/ComfyUI/LHM/ComfyUI-VideoHelperSuite /path/to/ComfyUI/custom_nodes/ComfyUI-VideoHelperSuite
# install packagescd LHM && bash install_cu118.sh &&cd ..
Prepare checkpoints
# in the source dir of ComfyUIcd models/checkpoints
wget https://virutalbuy-public.oss-cn-hangzhou.aliyuncs.com/share/aigc3d/data/LHM/ComfyUI/LHM_ComfyUI.zip
unzip LHM_ComfyUI.zip
rm LHM_ComfyUI.zip
cd ../..
# start the ComfyUI
python main.py
How to use
we provide full pipeline or you can infer from offline motion. We also provide a offline motion extraction pipeline.
Open LHM_Motion_Extraction.json, the extracted motion path will be shown after executation. This workflow is designed to avoid extracting motion params each time when you want to make videos based on the same motion. It will save much time in such case.
Animation from offline extracted motion (You only need 20s to generate a 10s video!!!)
Open LHM_Animation_Infer.json, paste specific offline motion path, upload your human image and click run.
Full Pipeline
Click 'Workflow' on the ComfyUI-FrondEnd and open LHM_WorkFlow.json, Upload your input image and driven video and click 'Run'!
About
[ICCV2025] LHM: Large Animatable Human Reconstruction Model from a Single Image in Seconds