• Upgraded to Torch 2.2.0.
• Changed the logic so that if a source image only contains a single face, and the target image contains multiple faces, only the first face in the target image is swapped with the face from the source image (previously the face from the source image would be swapped into all of the faces in the target image).
• Improved logging for the various different types of logic.
• Added installation script for Network Volumes.
• Updated installation instructions for Network Volumes.
• Attempted to add CUDAExecutionProvider to insightface FaceAnalysis but unfortunately its still not using it.
• Updated RunPod SDK to 1.6.0.

• Improved error handling.
• Refresh worker when there is an exception.
• Bumped to latest RunPod SDK version.
• Changed CMD to ENTRYPOINT to suppress spammy Nvidia notice.

• Improved error handling.

• Added support for providing specific face indexes of the faces to swap in the source and target images.
• Added support for either enabling or disabling face restoration.
• Added support for either enabling or disabling face upsampling.
• Added support for either enabling or disabling background enhancement.
• Added support for specifying an upscale value.
• Added support for specifying CodeFormer fidelity.
• Added support for specifying the output format (JPEG/PNG).

• Support swapping multiple faces from a single source image into multiple faces in a target image.

• Added support for swapping all faces in a target image that contains multiple faces with the image in the source image.
• Added support for providing target_index in the input field to specify the index of the face (left to right) that should be swapped in the target image if the target image contains multiple faces, where 0 is the first face, 1 is the second face, and so on.
• Restructured the documentation, improved the examples and added a list of RunPod statuses.

• Added support for either using a Network Volume or Standalone Docker image.

Initial release.