ComfyUI Node: KSampler Scheduled Sequence

KSampler Scheduled Sequence:

SaltKSamplerSequence is a powerful node designed to facilitate advanced sampling techniques in AI art generation. This node allows you to create intricate and detailed images by leveraging a sequence of sampling steps, conditioning inputs, and noise injection. It is particularly useful for generating high-quality latent images by iteratively refining the output through multiple stages. The node supports various configurations, including different samplers and schedulers, to provide flexibility and control over the sampling process. By using sequences for parameters like seed, denoise, and noise strength, SaltKSamplerSequence enables you to achieve nuanced and dynamic results, making it an essential tool for AI artists looking to push the boundaries of their creative projects.

KSampler Scheduled Sequence Input Parameters:

model

This parameter specifies the model to be used for sampling. It is a required input and determines the underlying architecture and capabilities of the sampling process.

seed_sequence

A list of integers used to initialize the random number generator for each sampling step. This sequence ensures reproducibility and variation in the generated images. Each seed in the sequence corresponds to a specific step in the sampling process.

steps

An integer that defines the number of sampling steps to be performed. The minimum value is 1, and the maximum value is 10000. The default value is 20. More steps generally lead to higher quality images but increase computation time.

cfg

A floating-point value representing the classifier-free guidance scale. It controls the trade-off between adhering to the conditioning inputs and the model's prior. The default value is 8.0, with a minimum of 0.0 and a maximum of 100.0.

sampler_name

Specifies the name of the sampler to be used. This parameter allows you to choose from different sampling algorithms provided by the KSampler class.

scheduler

Defines the scheduler to be used for the sampling process. Different schedulers can affect the timing and progression of the sampling steps.

positive_sequence

A list of conditioning inputs that guide the model towards desired features in the generated image. Each element in the sequence corresponds to a specific step in the sampling process.

negative_sequence

A list of conditioning inputs that guide the model away from undesired features in the generated image. The size of this sequence must match the positive_sequence.

latent_image

The initial latent image to be refined through the sampling process. This input serves as the starting point for the iterative refinement.

use_latent_interpolation

A boolean flag indicating whether to use latent interpolation during the sampling process. This can help in achieving smoother transitions between steps.

latent_interpolation_mode

Specifies the mode of latent interpolation to be used if use_latent_interpolation is enabled. Different modes can produce varying effects on the final image.

latent_interp_strength_sequence

A list of floating-point values representing the strength of latent interpolation at each step. This sequence allows for dynamic control over the interpolation process.

unsample_latents

A boolean flag indicating whether to perform unsampling on the latent images. Unsampling can help in refining the details of the generated image.

denoise_start

A floating-point value that specifies the starting point for denoising. This parameter helps in controlling the denoising process to achieve the desired level of detail.

denoise_sequence

A list of floating-point values representing the denoising strength at each step. This sequence allows for dynamic control over the denoising process.

inject_noise

A boolean flag indicating whether to inject noise into the latent images during the sampling process. Injecting noise can add variability and complexity to the generated images.

noise_strength_sequence

A list of floating-point values representing the strength of noise injection at each step. This sequence allows for dynamic control over the noise injection process.

KSampler Scheduled Sequence Output Parameters:

LATENT

The output of the SaltKSamplerSequence node is a refined latent image. This image has undergone multiple stages of sampling, conditioning, and noise injection to achieve a high level of detail and quality. The latent image can be further processed or used as the final output for AI art projects.

KSampler Scheduled Sequence Usage Tips:

• Experiment with different seed sequences to explore a variety of generated images and find the most appealing results.
• Adjust the number of steps based on the desired quality and available computational resources. More steps generally yield better results but require more processing time.
• Use the positive and negative sequences to fine-tune the features of the generated image. Ensure that both sequences are of the same length to avoid errors.
• Enable latent interpolation and adjust the interpolation strength sequence to achieve smoother transitions and more cohesive images.
• Inject noise at different strengths to add variability and complexity to the generated images. This can help in creating more dynamic and interesting results.

KSampler Scheduled Sequence Common Errors and Solutions:

ValueError: negative_sequence of size X does not match positive_sequence of size Y. Conditioning sizes must be the same.

• Explanation: This error occurs when the lengths of the positive_sequence and negative_sequence do not match.
• Solution: Ensure that both sequences are of the same length before running the node.

TypeError: Expected list for seed_sequence but got <type>

• Explanation: This error occurs when the seed_sequence is not provided as a list.
• Solution: Provide the seed_sequence as a list of integers.

IndexError: list index out of range

• Explanation: This error occurs when one of the sequences (denoise_sequence, latent_interp_strength_sequence, or noise_strength_sequence) is shorter than the number of steps.
• Solution: Use the expand_sequence method to ensure all sequences are of the appropriate length.

RuntimeError: Model not specified

• Explanation: This error occurs when the model parameter is not provided.
• Solution: Ensure that the model parameter is specified before running the node.