jieg9341-lab GitHub avatar

ComfyUI-Krea2-StyleTransfer

jieg9341-lab

A ComfyUI node that injects reference-image style signals into Krea2 sampling without training, enabling training-free style transfer with low content leakage and minimal quality loss.

Stars

174

7-day growth

No data

Forks

7

Open issues

0

License

MIT

Last updated

2026-07-21

AI repository intelligence
FR-AI / ANALYSIS

Why it is worth attention

It solves a practical gap in open-source Krea2 by providing a local style-reference implementation with a novel `ref_k_strength` control that decouples style transfer from content leakage, offering a tuned low-leakage preset.

Who it is for

  • ComfyUI users working with Krea2 models
  • AI artists seeking training-free style transfer
  • Developers experimenting with reference-injection techniques
  • Users who want to avoid official API dependencies

Use cases

  • Transfer a reference image's visual style to a new prompt without content leakage
  • Quickly prototype style variations without training a LoRA or preparing datasets
  • Combine two reference styles in an ordered, stage-based blend for controlled experimentation
  • Use as a temporary style adapter in local Krea2 workflows

Strengths

  • Independent `ref_k_strength` parameter enables low-leakage, high-quality style transfer by decoupling style signal from `low_scale_end`
  • Recommended single-reference mode provides a tested preset with strong style transfer and minimal quality loss
  • Simplified UX with `recommended` and `custom` modes for both normal users and experimenters
  • Two-reference mode offers explicit `primary_reference` and `first_phase_ratio` controls for ordered style blending

Considerations

  • Not the official Krea style reference module and does not replicate official multi-reference behavior
  • Two-reference transfer is experimental, order-sensitive, and stage-ratio-sensitive, with weaker results beyond two references
  • Very weak or generic reference images may not produce a strong style signal, and reference/target latent sizes must match

README quick start

Installation

Description

Training-free Krea2 style reference nodes for ComfyUI, with LoRA-like single-image transfer, experimental two-reference blending, low content leakage, and preserved image quality.

Related repositories

Similar projects matched by category, topics, and programming language.

lopopolo
Featured
lopopolo GitHub avatar

harness-engineering

Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.

AI & Machine LearningAI Agents
2,390
slvDev
Featured
slvDev GitHub avatar

esp32-ai

A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.

AI & Machine LearningLarge Language Models
1,960
littledivy
Featured
littledivy GitHub avatar

mimic

mimic captures traffic from any iOS or web app and automatically generates a Python client library that lets you call the app's API like a regular library.

AI & Machine Learning
1,482