ComfyUI Node: AdaptiveGuider

AdaptiveGuider:

AdaptiveGuidance is a specialized node designed to enhance the quality and precision of AI-generated art by dynamically adjusting the guidance scale during the sampling process. This node leverages adaptive guidance techniques to fine-tune the influence of conditioning inputs, ensuring that the generated outputs closely align with the desired artistic vision. By monitoring the similarity between positive and negative conditioning, AdaptiveGuidance can automatically adjust the classifier-free guidance (CFG) scale, providing a more controlled and refined output. This adaptive approach helps in achieving higher fidelity and more coherent results, making it an invaluable tool for AI artists seeking to produce high-quality, customized artwork.

AdaptiveGuider Input Parameters:

model

This parameter represents the AI model used for generating the artwork. It is a required input and serves as the backbone for the entire guidance process, ensuring that the adaptive adjustments are applied to the correct model.

positive

This parameter is a conditioning input that guides the model towards the desired output. It is a required input and significantly influences the final generated artwork by providing positive reinforcement to the model.

negative

This parameter is another conditioning input that guides the model away from undesired outputs. It is a required input and helps in refining the generated artwork by providing negative reinforcement to the model.

threshold

This parameter is a floating-point value that sets the similarity threshold for adjusting the CFG scale. The default value is 0.990, with a minimum of 0.90 and a maximum of 1.0. It determines how closely the positive and negative conditioning should match before the CFG scale is adjusted.

cfg

This parameter is a floating-point value that sets the initial classifier-free guidance scale. The default value is 8.0, with a minimum of 0.0 and a maximum of 100.0. It controls the overall influence of the conditioning inputs on the generated output.

uncond_zero_scale

This optional parameter is a floating-point value that sets the scale for zeroing out unconditional predictions. The default value is 0.0, with a maximum of 2.0. It helps in fine-tuning the balance between conditional and unconditional predictions.

cfg_start_pct

This optional parameter is a floating-point value that sets the starting percentage for the CFG scale adjustment. The default value is 0.0, with a maximum of 1.0. It determines the point in the sampling process where the CFG scale begins to adapt.

AdaptiveGuider Output Parameters:

GUIDER

This output parameter represents the configured guider object that applies the adaptive guidance during the sampling process. It encapsulates all the settings and adjustments made based on the input parameters, ensuring that the generated artwork adheres to the specified guidance.

AdaptiveGuider Usage Tips:

• To achieve more precise and high-quality results, experiment with different threshold values to find the optimal similarity level for your specific use case.
• Adjust the cfg parameter to control the overall influence of the conditioning inputs. Higher values can lead to more pronounced effects, while lower values provide subtler guidance.
• Utilize the uncond_zero_scale parameter to fine-tune the balance between conditional and unconditional predictions, especially when dealing with complex or ambiguous conditioning inputs.

AdaptiveGuider Common Errors and Solutions:

"Model object not found"

• Explanation: This error occurs when the specified model is not available or incorrectly referenced.
• Solution: Ensure that the model parameter is correctly specified and that the model is available in the expected location.

"Invalid threshold value"

• Explanation: This error occurs when the threshold value is outside the acceptable range.
• Solution: Verify that the threshold value is between 0.90 and 1.0 and adjust it accordingly.

"CFG scale out of range"

• Explanation: This error occurs when the cfg value is outside the acceptable range.
• Solution: Ensure that the cfg value is between 0.0 and 100.0 and adjust it as needed.

"Uncond zero scale out of range"

• Explanation: This error occurs when the uncond_zero_scale value exceeds the maximum limit.
• Solution: Verify that the uncond_zero_scale value is between 0.0 and 2.0 and adjust it accordingly.

"CFG start percentage out of range"

• Explanation: This error occurs when the cfg_start_pct value exceeds the maximum limit.
• Solution: Ensure that the cfg_start_pct value is between 0.0 and 1.0 and adjust it as needed.