inswapper Documentation

The face swap model must be placed in a ./checkpoints directory. For high-quality results, it is recommended to use CodeFormer for face restoration. The required models for CodeFormer will be downloaded automatically during the first inference run if you clone the repository as shown below.

mkdir checkpoints
wget -O ./checkpoints/inswapper_128.onnx https://github.com/facefusion/facefusion-assets/releases/download/models/inswapper_128.onnx

cd ..
git lfs install
git clone https://huggingface.co/spaces/sczhou/CodeFormer

To set up the environment, clone the repository, create a Python virtual environment, and install the required dependencies.

Note on Hardware Acceleration:

• For CPU-only inference: The default installation via requirements.txt will use onnxruntime.
• For GPU inference: You must manually install onnxruntime-gpu after following the standard installation steps.

# git clone this repository
git clone https://github.com/haofanwang/inswapper.git
cd inswapper

# create a Python venv
python3 -m venv venv

# activate the venv
source venv/bin/activate

# install required packages
pip install -r requirements.txt

You can use the process function from the swapper module to perform face swapping.

Parameters:

• source_img: A list of PIL Image objects representing the source faces.
• target_img: A PIL Image object representing the target image.
• model: Path to the ./checkpoints/inswapper_128.onnx model file.
• Additional integer parameters (e.g., -1, -1) are used for internal processing logic.

from swapper import *

source_img = [Image.open("./data/man1.jpeg"), Image.open("./data/man2.jpeg")]
target_img = Image.open("./data/mans1.jpeg")

model = "./checkpoints/inswapper_128.onnx"
result_image = process(source_img, target_img, -1, -1, model)
result_image.save("result.png")

The CLI allows for advanced face swapping including restoration, background enhancement, and upscaling.

CLI Arguments:

• --source_img: Semicolon-separated list of paths to source images (e.g., ./path1.jpg;./path2.jpg).
• --target_img: Path to the target image.
• --face_restore: Enables face restoration (using CodeFormer).
• --background_enhance: Enables background enhancement.
• --face_upsample: Enables face upsampling.
• --upscale: Scaling factor for upsampling (e.g., 2).
• --codeformer_fidelity: Fidelity setting for CodeFormer (e.g., 0.5).

python swapper.py \
--source_img="./data/man1.jpeg;./data/man2.jpeg" \
--target_img "./data/mans1.jpeg" \
--face_restore \
--background_enhance \
--face_upsample \
--upscale=2 \
--codeformer_fidelity=0.5