iFayens

iFayens / ComfyUI-PuLID-Flux2

Public

First PuLID implementation for FLUX.2 Klein (4B/9B) — Consistent face identity in ComfyUI

comfyui comfyui-nodes consistent-characters face-identity flux-klein
25
0
69% credibility
Found Mar 16, 2026 at 25 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Custom nodes for ComfyUI that adapt PuLID for face identity consistency in Flux.2 Klein image generation, including installation guides, workflows, and training scripts.

How It Works

1
🔍 Discover face-matching magic

You hear about a handy tool that lets your AI pictures keep the same face from a reference photo every time.

2
📥 Add to your AI art space

You copy the tool into your favorite AI image maker's extras folder so it's ready to use.

3
⬇️ Grab helper files

You download special face finder pieces and matching guides, placing them in the right spots like building blocks.

4
🧩 Set up face tools

You load the face finder, visual matcher, and guide file into your creation setup.

5
✨ Pick your reference face

You choose a photo of the face you love and connect it—now your AI knows exactly who to draw!

6
🎨 Create with consistency

You describe your scene, adjust the strength of the face match, and hit generate.

🥳 Perfect matching faces

Your AI art comes alive with the exact same face in every picture, just as you imagined.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 25 to 25 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is ComfyUI-PuLID-Flux2?

This Python custom node pack for ComfyUI delivers the first PuLID implementation tailored to Flux.2 Klein 4B/9B models, locking in consistent face identity across generations—like preserving a reference face's likeness in every output. Users drag in example workflows, load InsightFace detectors with buffalo_l fallbacks, EVA-CLIP encoders, and PuLID weights, then apply nodes to patch Flux models for identity-guided diffusion. Training scripts prep datasets from CelebA and train native weights, marking your first GitHub push toward custom models.

Why is it gaining traction?

It's the first GitHub repository adapting PuLID's architecture for Flux.2 Klein's transformer blocks, outpacing Flux.1-only alternatives with Klein-specific nodes and shared modulation support. Devs hook on the low-weight combo with reference conditioning for subtle consistency without quality dips, plus debug previews and auto-cleanup for green artifacts—perfect for your first GitHub contribution via trained weights.

Who should use this?

ComfyUI power users generating character art with Flux.2 Klein, like indie game devs needing consistent faces for NPCs or AI portrait artists tired of seed-tweaking. Suited for those opening their first GitHub issue on training tweaks or dataset preps.

Verdict

Grab it for Flux.2 Klein's first face consistency boost in ComfyUI—nodes run smoothly in beta v0.1—but with 19 stars and 0.7% credibility score, expect partial results until community-trained weights land. Solid docs and workflows make it worth a test run over Flux.1 PuLID forks.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.