This repository provides a ComfyUI implementation of PhotoMaker, a tool for customizing realistic human photos using stacked ID embeddings. It's designed for users of ComfyUI, Stable Diffusion's node-based interface, offering enhanced face resemblance and flexibility through customizable trigger words and additional nodes for style and image preparation.

How It Works

The implementation leverages InsightFace for its PhotoMaker V2 support, integrating with ComfyUI's node-based workflow. It utilizes CLIPImageProcessor for improved face resemblance, similar to the original PhotoMaker code. The design prioritizes memory efficiency and speed, aiming for stability across ComfyUI updates.

Quick Start & Requirements

• Install ComfyUI.
• Install onnxruntime and insightface.
• Clone the repository into ComfyUI/custom_nodes.
• Download PhotoMaker models (V1, V2) from Hugging Face and place them in ComfyUI/models/photomaker.
• Example workflows are available.

Highlighted Details

• Enhanced face resemblance via CLIPImageProcessor.
• Customizable trigger words and support for multiple trigger words.
• Additional nodes: PhotoMakerStyles and PrepImagesForClipVisionFromPath.
• Added PhotoMakerLoraLoaderPlus node for LoRA loading.
• Support for PhotoMaker V2 using InsightFace.

Maintenance & Community

The project has seen recent updates, including support for PhotoMaker V2 and a new LoRA loader node. The repository name was changed to reflect official PhotoMaker integration into ComfyUI.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

Users may need to recreate nodes due to recent updates. The ref_images_path input has been removed in favor of a new node.