PuLID ComfyUI is a custom nodes repository designed to integrate the PuLID framework into ComfyUI. This integration enables users to directly implement PuLID's capabilities within ComfyUI workflows. PuLID, established as a native implementation, leverages advanced machine learning models to improve image generation fidelity and style adaptation.
To successfully install the PuLID ComfyUI custom nodes, follow these steps:
Download the PuLID pre-trained model:
ComfyUI/models/pulid/
.Automatic Download of EVA CLIP:
Install facexlib
Dependency:
facexlib
is installed. The necessary models will be downloaded on first use.Install InsightFace with AntelopeV2:
ComfyUI/models/insightface/models/antelopev2
.This repository provides a set of custom nodes that facilitate the integration of the PuLID framework within ComfyUI. This integration focuses on enhancing image generation processes by improving attention masking, style fidelity, and flexibility to align closely with reference images or styles.
This repository offers the following nodes:
Attention Masking and Advanced Node: Added on 2024.05.12, these features enhance the ability to fine-tune image generation by adjusting how closely the generated image adheres to a reference.
Method Parameter: Allows different approaches in applying weights:
Fidelity
: Ensures resemblance to the reference ID.Style
: Offers more style freedom.Neutral
: Provides no normalization, with adjustments applied manually in advanced settings.Projection and Fidelity Options in the Advanced Node: Facilitates nuanced control over the stylistic and fidelity outcomes of the generated images.
PuLID ComfyUI can be instrumental in workflows that require high-level image manipulation, particularly:
By embedding these nodes into their workflows, ComfyUI users can achieve more precise and stylistically coherent results, making it an invaluable tool for artists and developers working within this UI framework.