This repository provides an unofficial implementation of the PhotoMaker model for ComfyUI, enabling users to generate images with consistent identity across various styles and poses. It targets users of ComfyUI, particularly those interested in character consistency for creative image generation, offering enhanced control and flexibility over the original PhotoMaker.

How It Works

The implementation integrates PhotoMaker as custom nodes within the ComfyUI ecosystem. It supports loading base SDXL models and PhotoMaker adapters from Hugging Face Hub or local paths. Key features include flexible reference image preprocessing (direct input or path-based), LoRA model loading with weight adjustment, and a Prompt_Styler node compatible with various text-to-image prompts and styles, including official PhotoMaker styles and custom ones.

Quick Start & Requirements

• Install: Navigate to custom_nodes, clone the repository (git clone https://github.com/ZHO-ZHO-ZHO/ComfyUI-PhotoMaker.git), then cd custom_nodes/ComfyUI-PhotoMaker and run pip install -r requirements.txt. Restart ComfyUI.
• Prerequisites: SDXL compatible models, Python.
• Resources: No specific hardware requirements beyond standard ComfyUI setup are listed, but performance claims suggest GPU acceleration is highly beneficial.
• Links: Video Demo: https://github.com/ZHO-ZHO-ZHO/ComfyUI-PhotoMaker/assets/140084057/d58af6e7-d0f3-41ff-ab33-195cb6d66e9e.

Highlighted Details

• Supports LoRA integration and batch generation.
• Compatible with generic prompt stylers like Styler and Portrait Master.
• Offers 10 built-in styles plus custom style support.
• Claims a 3x speed increase over previous versions (e.g., 7s on A100 for 50 steps).

Maintenance & Community

The project is actively updated, with recent additions including Disney-character workflows and LoRA/batch support. A QQ group (839821928) is available for community interaction.

Licensing & Compatibility

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

Limitations & Caveats

The implementation is unofficial. While it supports custom dimensions, the core PhotoMaker generation requires 1024x1024 resolution. The project is under active development, with workflows for V2.0 being deprecated in favor of V2.5.