### Download LivePortrait Installer from HuggingFace

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-05.md

Use the huggingface-cli to download the Windows installer. Ensure huggingface_hub is installed.

```bash
# !pip install -U "huggingface_hub[cli]"
huggingface-cli download cleardusk/LivePortrait-Windows LivePortrait-Windows-v20240806.zip --local-dir ./
```

--------------------------------

### Download LivePortrait Installer via HF-Mirror

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-05.md

Download the Windows installer using huggingface-cli with HF_ENDPOINT set to hf-mirror.com. This is an alternative if direct Hugging Face access is restricted. Ensure huggingface_hub is installed.

```bash
# !pip install -U "huggingface_hub[cli]"
export HF_ENDPOINT=https://hf-mirror.com
huggingface-cli download cleardusk/LivePortrait-Windows LivePortrait-Windows-v20240806.zip --local-dir ./
```

--------------------------------

### Install FFmpeg on Ubuntu/Debian

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/how-to-install-ffmpeg.md

Commands for installing FFmpeg and its development libraries on Ubuntu or Debian-based systems.

```bash
sudo apt install ffmpeg
```

```bash
sudo apt install libsox-dev
```

```bash
conda install -c conda-forge 'ffmpeg<7'
```

--------------------------------

### Install Project Dependencies

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Install all remaining Python dependencies required for the LivePortrait project from the requirements.txt file.

```bash
pip install -r requirements.txt
```

--------------------------------

### Launch Gradio Interface (Linux/Windows)

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Start the Gradio web interface for human mode inference on Linux and Windows systems.

```bash
python app.py
```

--------------------------------

### Install FFmpeg on MacOS

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/how-to-install-ffmpeg.md

Use Homebrew to install FFmpeg on MacOS.

```bash
brew install ffmpeg
```

--------------------------------

### Launch Gradio Interface for Animals

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Start the Gradio web interface specifically for animal inference. This is tested on Linux with NVIDIA GPU.

```bash
python app_animals.py
```

--------------------------------

### Install PyTorch for CUDA 11.1

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Install PyTorch, torchvision, and torchaudio compatible with CUDA 11.1. This command uses a specific index URL for CUDA 11.1 wheels.

```bash
pip install torch==1.10.1+cu111 torchvision==0.11.2 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
```

--------------------------------

### Install FFmpeg with Conda

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/how-to-install-ffmpeg.md

Use this command if you are a Conda user to install FFmpeg.

```bash
conda install ffmpeg
```

--------------------------------

### Install PyTorch for CUDA 11.8

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Install PyTorch, torchvision, and torchaudio compatible with CUDA 11.8. This command uses the default PyTorch index URL for CUDA 11.8 wheels.

```bash
pip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu118
```

--------------------------------

### Install PyTorch for CUDA 12.1

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Install PyTorch, torchvision, and torchaudio compatible with CUDA 12.1. This command uses the default PyTorch index URL for CUDA 12.1 wheels.

```bash
pip install torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu121
```

--------------------------------

### Install Dependencies for macOS

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Install Python dependencies specifically for macOS with Apple Silicon, using a separate requirements file as X-Pose is not supported.

```bash
pip install -r requirements_macOS.txt
```

--------------------------------

### Launch Animals Mode Gradio App

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-02.md

Start the Gradio interface for the animals mode. Optional arguments can be used to specify server port, name, and sharing.

```bash
python app_animals.py # --server_port 8889 --server_name "0.0.0.0" --share
```

--------------------------------

### Clone Repository and Create Conda Environment

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Clone the LivePortrait repository and set up a dedicated Conda environment for the project. Ensure you have git, conda, and FFmpeg installed.

```bash
git clone https://github.com/KlingTeam/LivePortrait
cd LivePortrait

# create env using conda
conda create -n LivePortrait python=3.10
conda activate LivePortrait
```

--------------------------------

### Check CUDA Version

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Verify your system's CUDA version. This is important for selecting the correct PyTorch installation.

```bash
nvcc -V # example versions: 11.1, 11.8, 12.1, etc.
```

--------------------------------

### Display Inference Help

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

View all available command-line options for the inference script.

```bash
python inference.py -h
```

--------------------------------

### Run Inference with Video Source

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Perform inference using a video as the source input and another video as the driving input.

```bash
python inference.py -s assets/examples/source/s13.mp4 -d assets/examples/driving/d0.mp4
```

--------------------------------

### Download Pretrained Weights from HuggingFace

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Download pretrained weights from the HuggingFace model hub using the huggingface-cli. This command excludes Git-related files and documentation.

```bash
# !pip install -U "huggingface_hub[cli]"
huggingface-cli download KlingTeam/LivePortrait --local-dir pretrained_weights --exclude "*.git*" "README.md" "docs"
```

--------------------------------

### Launch Gradio with Torch Compile

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Enable torch.compile for faster inference via the Gradio interface. This optimization process takes about a minute on the first run and is not supported on Windows or macOS.

```bash
python app.py --flag_do_torch_compile
```

--------------------------------

### Launch Gradio Interface (macOS Apple Silicon)

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Run the Gradio web interface for human mode on macOS with Apple Silicon. Note potential performance limitations.

```bash
PYTORCH_ENABLE_MPS_FALLBACK=1 python app.py
```

--------------------------------

### Run Inference with Motion Template (Video Source)

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Edit a portrait video using a pre-generated motion template (.pkl file). This method can accelerate inference and enhance privacy.

```bash
python inference.py -s assets/examples/source/s13.mp4 -d assets/examples/driving/d5.pkl
```

--------------------------------

### Download Pretrained Weights via HF-Mirror

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Download pretrained weights using hf-mirror as an alternative endpoint for HuggingFace. Set the HF_ENDPOINT environment variable before downloading.

```bash
# !pip install -U "huggingface_hub[cli]"
export HF_ENDPOINT=https://hf-mirror.com
huggingface-cli download KlingTeam/LivePortrait --local-dir pretrained_weights --exclude "*.git*" "README.md" "docs"
```

--------------------------------

### Run Inference with Image Source

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Perform inference using a static image as the source input and a video as the driving input.

```bash
python inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d0.mp4
```

--------------------------------

### Run Inference with Motion Template (Image Source)

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Generate a portrait animation using a static image source and a pre-generated motion template (.pkl file). This can speed up inference and protect privacy.

```bash
python inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d5.pkl
```

--------------------------------

### Build X-Pose Attention Module

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Compile the MultiScaleDeformableAttention module required for animal inference. Ensure your CUDA versions are compatible.

```bash
cd src/utils/dependencies/XPose/models/UniPose/ops
python setup.py build install
cd - # equal to cd ../../../../../../../
```

--------------------------------

### Build X-Pose OP MultiScaleDeformableAttention

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-02.md

Build the MultiScaleDeformableAttention OP required for X-Pose keypoint detection. Ensure you are in the correct directory before running.

```bash
cd src/utils/dependencies/XPose/models/UniPose/ops
python setup.py build install
cd - # this returns to the previous directory
```

--------------------------------

### Run Human Inference (macOS Apple Silicon)

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Run the inference script on macOS with Apple Silicon, noting potential performance differences compared to NVIDIA GPUs.

```bash
PYTORCH_ENABLE_MPS_FALLBACK=1 python inference.py
```

--------------------------------

### Evaluate Inference Speed

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Run the script to benchmark the inference speed of each module on NVIDIA GPUs.

```bash
python speed.py
```

--------------------------------

### Use Image as Driving Reference (CLI)

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-19.md

Use this command to specify a driving image for animation. Ensure the source and driving files are correctly path-ed.

```bash
python inference.py -s assets/examples/source/s5.jpg -d assets/examples/driving/d30.jpg
```

--------------------------------

### Run Inference with Auto-Cropping

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Perform inference while enabling the auto-cropping feature for the driving video. This is recommended for custom driving videos.

```bash
python inference.py -s assets/examples/source/s9.jpg -d assets/examples/driving/d13.mp4 --flag_crop_driving_video
```

--------------------------------

### Run Animal Inference

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Execute inference for animal portraits using a specified source, driving input (PKL file), and custom multipliers. This mode is tested on Linux/Windows with NVIDIA GPU.

```bash
python inference_animals.py -s assets/examples/source/s39.jpg -d assets/examples/driving/wink.pkl --driving_multiplier 1.75 --no_flag_stitching
```

--------------------------------

### Run Human Inference (Linux/Windows)

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Execute the main inference script for human portraits on Linux and Windows systems.

```bash
python inference.py
```

--------------------------------

### Configure Animals Model Version

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2025-01-01.md

Modify this configuration to switch between the new (v1.1) and original Animals model versions. Ensure your weights are updated to use the new version.

```python
version_animals = "" # old version
# version_animals = "_v1.1" # new (v1.1) version
```

--------------------------------

### Run Animals Mode Inference

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-02.md

Execute the animals mode inference script. Disable stitching and adjust the driving multiplier as recommended for animal models.

```bash
python inference_animals.py -s assets/examples/source/s39.jpg -d assets/examples/driving/wink.pkl --no_flag_stitching --driving_multiplier 1.75
```

--------------------------------

### Regional Control for Animation (CLI)

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-19.md

Specify the animation region to control specific facial features. Use 'lip' to animate only the lip movements.

```bash
# only driving the lip region
python inference.py -s assets/examples/source/s5.jpg -d assets/examples/driving/d0.mp4 --animation_region lip
```

--------------------------------

### LivePortrait Citation BibTeX

Source: https://github.com/klingairesearch/liveportrait/blob/main/readme.md

Use this BibTeX entry to cite the LivePortrait project in your research or publications. It includes title, authors, journal, and year.

```bibtex
@article{guo2024liveportrait,
  title   = {LivePortrait: Efficient Portrait Animation with Stitching and Retargeting Control},
  author  = {Guo, Jianzhu and Zhang, Dingyun and Liu, Xiaoqiang and Zhong, Zhizhou and Zhang, Yuan and Wan, Pengfei and Zhang, Di},
  journal = {arXiv preprint arXiv:2407.03168},
  year    = {2024}
}
```

--------------------------------

### Absolute Motion Driving (CLI)

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-08-19.md

Use this command to apply the driving video's motion directly to the output, disabling relative motion. This may cause jitter or identity leakage.

```bash
python inference.py --no_flag_relative_motion
```

--------------------------------

### Adjusting Detection Threshold in LivePortrait

Source: https://github.com/klingairesearch/liveportrait/blob/main/assets/docs/changelog/2024-07-24.md

Use the `--det_thresh` argument to modify the face detection threshold. A lower value increases recall, detecting more varied faces. The default is 0.15.

```python
python app.py --det_thresh 0.5
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.