ComfyUI-Apply_Style_Model_Adjust

ComfyUI-Apply_Style_Model_Adjust
★ 10

风格迁移FLUX 模型提示与风格平衡自定义节点
为使用 FLUX 风格模型的风格迁移提供增强的权重平衡控制，提升降低风格强度时文本提示的影响力，平滑风格与提示的过渡。
💡 在使用 FLUX 风格模型时调节风格与提示的平衡。
🍴 3 Forks💻 Python🔄 2024-11-23
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/8f9eee5e2cdb
image
📄 README

Apply Style Model Adjust for ComfyUI

A custom node that provides enhanced control over style transfer balance when

using FLUX style models in ComfyUI. This node offers better control over the

influence of text prompts versus style reference images.

Features

  • Enhanced prompt influence when reducing style strength
  • Better balance between style transfer and text prompt conditioning
  • Smoother transition between style and prompt dominance
  • Specifically optimized for FLUX style models and image variations

    • Installation

  • Navigate to your ComfyUI’s custom_nodes directory
  • Clone this repository:
    • cd custom_nodes
git clone https://github.com/ShmuelRonen/ComfyUI-Apply_Style_Model_Adjust.git

    Or download and extract the zip file into your ComfyUI/custom_nodes folder.

    Usage

    The node appears as “Apply Style Model (Adjusted)” under the

    conditioning/style_model category.

    Node Parameters

  • conditioning: Base conditioning from your text prompt
  • style_model: Loaded FLUX style model
  • clip_vision_output: CLIP Vision encoding of reference image
  • strength: Balance between style and prompt (0.0 – 1.0)
  • Low values (0.0-0.3): Text prompt dominates
  • Medium values (0.4-0.7): Balanced mix
  • High values (0.8-1.0): Style reference dominates

    • Difference from Standard Style Model Apply

    The standard Style Model Apply node can sometimes:

  • Let style overwhelm prompt instructions
  • Have abrupt transitions when adjusting strength
  • Provide less control over the final result

    • This adjusted version:

  • Maintains better prompt influence
  • Provides smoother strength transitions
  • Reduces maximum style influence
  • Better handles text instructions while maintaining style

    • FLUX Model Note

    This node is optimized for use with FLUX.1 Redux which:

  • Is designed for image variation generation
  • Works as an adapter for FLUX.1 base models
  • Naturally integrates into complex workflows
  • Best used for subtle style adjustments

    • Requirements

  • ComfyUI (latest version)
  • FLUX style models
  • Standard ComfyUI dependencies

    • Known Limitations

  • Primary focus is on FLUX model compatibility
  • Very low strength values (<0.2) might affect image quality
  • Cannot override FLUX model’s inherent limitations:
  • Heavy influence from input image
  • Limited ability to add new elements
  • Focus on variations rather than major changes

    • Support

    If you encounter issues:

  • Open an issue on GitHub
  • Check existing issues first
  • Provide your:
  • ComfyUI version
  • Workflow setup
  • Error messages if any

    • Credits

    Based on ComfyUI’s original Style Model Apply node, with modifications for

    better control and balance.

    License

    MIT License – feel free to use, modify, and distribute.

    Version History

    v1.0

  • Initial release
  • Improved balance control
  • Better prompt influence scaling