ComfyUI Node: Pre CFG PAG

Pre CFG PAG:

The Pre CFG PAG node, short for Pre-Conditioned Perturbed Attention Guidance, is designed to enhance the quality and precision of AI-generated images by applying a sophisticated attention mechanism before the CFG (Classifier-Free Guidance) process. This node leverages perturbed attention guidance to refine the conditioning of the model, ensuring that the generated images are more aligned with the desired attributes and details specified by the user. By integrating this node into your workflow, you can achieve more accurate and visually appealing results, as it helps the model focus on the most relevant features during the generation process.

Pre CFG PAG Input Parameters:

model

This parameter specifies the model to which the Pre CFG PAG node will be applied. It is essential for defining the base model that will undergo the perturbed attention guidance process.

start_at_sigma

This parameter determines the starting point of the sigma range for applying the perturbed attention guidance. The sigma value influences the noise level in the model, and setting this parameter helps control when the guidance should begin. The default value is 15.0, with a minimum of 0.0 and a maximum of 1000.0.

end_at_sigma

This parameter sets the endpoint of the sigma range for the perturbed attention guidance. It defines when the guidance should stop, allowing for precise control over the duration of the process. The default value is 1.0, with a minimum of 0.0 and a maximum of 1000.0.

enabled

This boolean parameter enables or disables the perturbed attention guidance. When set to true, the guidance is applied; when false, it is bypassed. The default value is true.

Pre CFG PAG Output Parameters:

model

The output parameter is the model that has undergone the perturbed attention guidance process. This refined model is expected to produce more accurate and visually appealing images, as it has been conditioned to focus on the most relevant features during the generation process.

Pre CFG PAG Usage Tips:

• To achieve the best results, experiment with different values for start_at_sigma and end_at_sigma to find the optimal range for your specific use case.
• Ensure that the enabled parameter is set to true to activate the perturbed attention guidance and benefit from its enhancements.

Pre CFG PAG Common Errors and Solutions:

Model not specified

• Explanation: The model parameter is missing or not correctly specified.
• Solution: Ensure that you provide a valid model to the node before execution.

Invalid sigma range

• Explanation: The values for start_at_sigma and end_at_sigma are out of the acceptable range or incorrectly set.
• Solution: Verify that the sigma values are within the range of 0.0 to 1000.0 and that start_at_sigma is greater than end_at_sigma.

Guidance not applied

• Explanation: The enabled parameter is set to false, so the perturbed attention guidance is not applied.
• Solution: Set the enabled parameter to true to activate the guidance process.