ACN_ReferenceControlNet Node Documentation

Introduction

The ACN_ReferenceControlNet node is a part of the ComfyUI-Advanced-ControlNet repository, which offers advanced capabilities for scheduling ControlNet strength across timesteps and latents in machine learning workflows. This node specifically deals with the reference aspect of ControlNet, allowing users to integrate reference images and settings into their workflows for enhanced control over the image generation process.

Overview

The ACN_ReferenceControlNet node provides the ability to incorporate reference-based conditions into a ControlNet, giving users the flexibility to balance reference-influenced outcomes with standard outputs. It supports modes such as reference_attn, reference_adain, and reference_adain+attn, alongside options for style fidelity and reference weight adjustments.

Key Features

• Reference Modes: Supports multiple modes including reference_attn, reference_adain, and reference_adain+attn.
• Style Fidelity Control: Users can control the fidelity of style references similarly to Auto1111's features.
• Weighted Reference Influence: Adjusts the weight given to the reference image's influence on the output.
• Integration with Advanced ControlNet Features: Fully compatible with other advanced features provided by this repository.

Inputs

The node accepts the following inputs:

• conditioning (positive): Positive conditioning signals, typically part of the image generation process.
• conditioning (negative): Negative conditioning signals used to refine the output.
• loaded controlnet (converted to Advanced version): The controlnet to which the reference conditions are applied.
• reference image: Images used as a reference in guiding the controlnet, influencing the style and characteristics of the output.

Optional inputs might include:

• attention masks: Masks indicating the areas of the image where reference influence should be applied, and to what extent.
• weights: Additional settings to override default weightings for reference application.

Outputs

The node produces the following outputs:

• conditioning (positive) with applied controlnets: The positive conditioning signals after the reference conditions have been applied.
• conditioning (negative) with applied controlnets: The negative conditioning signals post-application of the reference conditions.

Usage in ComfyUI Workflows

The ACN_ReferenceControlNet node is used within ComfyUI workflows to enhance the influence of reference images on the generated outputs. By employing this node, users can fine-tune the control and impact of reference images, ensuring the resulting outputs align more closely with desired styles or characteristics.

Typical Workflow Inclusions

1. Loading Reference Images: Incorporate reference images that guide the ControlNet's influence over the generated output.
2. Applying Attention Masks: Use masks to control where and how strongly the reference image should influence different parts of the output.
3. Adjusting Weights: Fine-tune the balance of reference influence on the generated imagery to achieve desired fidelity and style.
4. Integration with Other Nodes: Combine with other nodes such as Apply Advanced ControlNet for more complex scheduling and strength adjustments.

Special Features and Considerations

• Modes and Customization: The node supports several modes for applying reference conditions, enabling users to customize the approach based on their needs.
• Balance Control: By setting style_fidelity and ref_weight, users can adjust the extent to which reference images influence the ControlNet, allowing for fine-tuning between reference and no-reference results.
• Compatibility: Fully compatible with other advanced nodes in the repository, allowing for seamless incorporation into more complex machine learning workflows.

In conclusion, the ACN_ReferenceControlNet node is a versatile and powerful feature of the ComfyUI-Advanced-ControlNet suite, offering users significant control over how reference images are integrated into ControlNet functionalities, ultimately influencing the form and style of generated outputs significantly.