ComfyUI Node: StructuredNoise

StructuredNoise:

The StructuredNoise node is designed to generate structured noise that preserves the phase of an image while introducing controlled randomness. This node is particularly useful in applications where maintaining the structural integrity of an image is crucial, such as in phase-preserving diffusion processes. By combining Gaussian noise magnitudes with the image's phase information, the node allows for a nuanced manipulation of noise, ensuring that low-frequency components retain their original phase while high-frequency components adopt a noise phase. This approach provides a sophisticated method for noise generation that can enhance the realism and detail in AI-generated art, making it a valuable tool for artists looking to add texture and complexity to their creations.

StructuredNoise Input Parameters:

latent

The latent parameter serves as the input tensor that represents the initial state or structure upon which the noise will be applied. It is crucial for defining the dimensions and the base characteristics of the noise to be generated. This parameter does not have a specified range as it depends on the specific application and the dimensions of the input data.

seed

The seed parameter is an integer that initializes the random number generator, ensuring reproducibility of the noise pattern. By setting a specific seed value, you can generate the same noise pattern across different runs. The minimum value is 0, and the maximum is 0xffffffffffffffff, with a default value of 0.

noise_std

The noise_std parameter controls the standard deviation of the Gaussian noise applied to the image. It determines the intensity of the noise, with higher values resulting in more pronounced noise. The minimum value is 0.0, the maximum is 10.0, and the default is 1.0.

cutoff_radius

The cutoff_radius parameter defines the frequency cutoff radius, which determines the extent to which the original image structure is preserved. A value of 0 means full structure preservation, while higher values allow more noise to alter the image. The range is from 0.0 to 1000.0, with a default of 40.0.

transition_width

The transition_width parameter specifies the width of the transition between the preserved image phase and the noise phase. It affects how smoothly the noise blends with the image, with larger values resulting in a more gradual transition. The minimum value is 0.1, the maximum is 20.0, and the default is 2.0.

pad_factor

The pad_factor parameter determines the padding factor applied to the image before noise generation. This helps in managing edge effects during the Fourier transform process. The range is from 1.0 to 3.0, with a default value of 1.5.

sampling_method

The sampling_method parameter specifies the method used for sampling the noise. This parameter allows for different techniques to be applied, which can affect the characteristics of the generated noise. The specific options for this parameter are not detailed in the context.

input_noise_tensor

The input_noise_tensor parameter provides an optional input tensor for noise, which can be used to introduce pre-defined noise patterns into the process. This allows for greater control and customization of the noise characteristics.

StructuredNoise Output Parameters:

noise_obj

The noise_obj is the output parameter that contains the generated structured noise. This object encapsulates the noise pattern that has been applied to the latent input, incorporating the specified parameters such as noise standard deviation, cutoff radius, and transition width. The output is crucial for further processing or direct application in AI art projects, providing a noise pattern that maintains the structural integrity of the original image while introducing controlled randomness.

StructuredNoise Usage Tips:

• Experiment with the cutoff_radius to balance between preserving the original image structure and introducing new noise patterns. Lower values will maintain more of the original image's characteristics.
• Use the seed parameter to ensure consistency across different runs, especially when you need to reproduce specific noise patterns for comparison or iterative design processes.
• Adjust the noise_std to control the intensity of the noise. Start with the default value and gradually increase it to see how it affects the image, especially in terms of texture and detail.

StructuredNoise Common Errors and Solutions:

ImportError: Could not find structured_noise package.

• Explanation: This error occurs when the structured_noise package is not installed or not found in the expected directory.
• Solution: Ensure that the structured_noise package is installed by running pip install structured-noise. If the package is cloned into a custom directory, make sure the path is correctly set in your environment.

ValueError: Invalid parameter range.

• Explanation: This error might occur if an input parameter is set outside its allowed range.
• Solution: Double-check the parameter values to ensure they fall within the specified minimum and maximum limits. Adjust the values accordingly and try again.