Clone this repo:
git clone https://github.com/neuralchen/SimHuman.git
cd SimHumanAll dependencies for defining the environment are provided in environment.yml. We recommend running this repository using Anaconda (you may need to modify environment.yml to install PyTorch that matches your own CUDA version following https://pytorch.org/):
conda env create -f environment.ymlTensorRT download URL: https://developer.nvidia.com/tensorrt-download
CUDA version must match with TensorRT version.
Untar TensorRT installation package:
tar xzvf TensorRT-xxxxxxxxxxxx.tarAdd system environment PATH in /etc/profile or ~/.bashrc
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:[path to your tensorrt]/libCopy files to system lib:
sudo cp -r [your tensorrt dir]/lib/* /usr/lib
sudo cp -r [your tensorrt dir]/include/* /usr/includeInstallation of TensorRT, Before installation, please check the python version:
# TensorRT installation
cd [your tensorrt dir]/python
pip install tensorrt-xxxxxxxx-py2.py3-none-any.whl
# UFF installation to support model conversion of tensorflow
cd [your tensorrt dir]/uff
pip install uff-xxxxxxxx-py2.py3-none-any.whl
# graphsurgeon installation, to support self-defined operators and structures
cd [your tensorrt dir]/graphsurgeon
pip install graphsurgeon-xxxxxxxxxx-py2.py3-none-any.whlTo verify installation, please import tensorrt and uff in python env.
Download face detection models from Google Driver.
- glintr100.onnx ---> ./insightface_func/models/ckpt/antelope/
- scrfd_10g_bnkps.onnx ---> ./insightface_func/models/ckpt/antelope/
- det_10g.onnx ---> ./insightface_func/new/
Download TTS models from [Baidu Driver, Password：qgpi]
- 8000.pth.tar ---> ./TTS/output/ckpt/biaobei/
- generator_universal.pth.tar ---> ./TTS/hifigan/
- best_model.pt ---> ./TTS/transformer/prosody_model/
- GFPGANv1.4.onnx ---> ./restoration/
- wav2lip.pth ---> ./checkpoints/
python gui.py --video ./example.mp4