Deploying DiffPortrait360 requires a step-by-step process to complete the environment preparation:
Hardware preparation
- Must use NVIDIA GPU (CUDA support)
- Minimum configuration: 30GB video memory (generates 32 fps)
- Recommended configuration: 80GB video memory (e.g. A6000 graphics card)
software installation
- Confirmation of CUDA version (recommended 12.2): passed
nvcc --versionprobe - Create a Conda environment for Python 3.9:
conda create -n diffportrait360 python=3.9conda activate diffportrait360
- Clone the code repository:
git clone https://github.com/FreedomGu/DiffPortrait360.git - Install dependencies:
- go into
diffportrait360_releasecatalogs - fulfillment
pip install -r requirements.txt
- go into
Model deployment
Download the three pre-trained models (PANO_HEAD_MODEL/Head_Back_MODEL/Diff360_MODEL) from Hugging Face and modify theinference.shThe path parameters in the Finally, the path parameter is passed through thetorch.cuda.is_available()Verify GPU availability.
This answer comes from the articleDiffPortrait360: Generate 360-degree head views from a single portraitThe
May not be reproduced without permission:AI productivity tools " How to install and configure DiffPortrait360 runtime environment?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil