shootthesound

shootthesound / comfyui-ReferenceLatentPlus

Public

Improved ReferenceLatent for ComfyUI — per-image strength, timestep gating, MediaPipe auto-masks, 1-4 image inputs. Works with Flux, Klein, Lumina, Wan, Hunyuan, Qwen-Image.

ai comfyui flux generative-ai hunyuan
18
2
100% credibility
Found May 13, 2026 at 18 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Custom node for ComfyUI that enhances reference image guidance in AI generation with adjustable strength, masking, multiple inputs, and size limits.

How It Works

1
🔍 Find the helpful tool

While exploring ways to make your AI art match reference photos better, you discover this handy add-on for your picture-making program.

2
📥 Add it to your setup

Download the folder and drop it into your picture program's custom tools folder.

3
🔄 Wake it up

Restart your picture program, and the new Reference Latent+ tool appears right in the menu.

4
🖼️ Load your favorite photos

Pick one to four reference pictures—like a face or scene—and connect them easily to the tool.

5
⚙️ Customize the influence

Tweak sliders for how strongly each photo guides the result, choose parts like face or body to focus on, and decide when it kicks in.

6
🚀 Generate magic

Connect it to your text description and press go—watch as it blends your photos perfectly into new creations.

🎉 Stunning results

Celebrate your beautiful AI images that look just like you wanted, with precise control from your references.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 18 to 18 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-ReferenceLatentPlus?

This Python node for ComfyUI upgrades the basic ReferenceLatent with per-image strength controls, timestep gating, MediaPipe auto-masks for face/body/clothes/background, and slots for 1-4 image inputs. It solves the stock node's limits—no fine-tuned strength, full-image bleed, manual multi-ref wiring, or compute blowup from huge refs—letting you feed precise reference images into Flux, Klein, Lumina, Hunyuan, and similar diffusion models without workflow spaghetti.

Why is it gaining traction?

Unlike the binary stock version, it offers signed strength (-5 to 50), per-ref timestep ranges to layer influences (e.g., style late, pose early), and novel mask fills like latent zero or noise for cleaner isolation. The megapixel cap keeps attention costs in check, auto-masks save mask-drawing time, and progressive inputs reveal as you connect—streamlining ComfyUI workflows for improved image refs without extra nodes.

Who should use this?

ComfyUI power users generating consistent characters or styles in Flux/Klein/Lumina/Hunyuan pipelines, especially those battling ref leakage or multi-image compositing. Ideal for artists iterating on identity transfer, pose guidance, or environment refs where stock tools force hacks like chained encoders.

Verdict

Grab it if you hit ReferenceLatent walls—docs and example workflows make setup fast, and it works out-of-box on supported models. At 18 stars and 1.0% credibility, it's early but battle-tested on Klein 9B; test in non-prod first as masking tweaks are novel.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.