ComfyUI > Nodes > CRT-Nodes > Latent Noise Injection Sampler (CRT)

ComfyUI Node: Latent Noise Injection Sampler (CRT)

Class Name

LatentNoiseInjectionSampler

Category
CRT/Sampling
Author
CRT (Account age: 1707days)
Extension
CRT-Nodes
Latest Updated
2026-03-16
Github Stars
0.1K

How to Install CRT-Nodes

Install this extension via the ComfyUI Manager by searching for CRT-Nodes
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter CRT-Nodes in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

Latent Noise Injection Sampler (CRT) Description

Enhances image generation by injecting adjustable noise into latent space for diverse outputs.

Latent Noise Injection Sampler (CRT):

The LatentNoiseInjectionSampler is a sophisticated node designed to enhance the sampling process by injecting noise into the latent space during image generation. This node allows you to introduce controlled randomness into the sampling process, which can lead to more diverse and creative outputs. By enabling noise injection, you can explore a wider range of artistic possibilities, as the node modifies the latent space with noise that can be adjusted in terms of strength and timing. The primary goal of this node is to provide a flexible tool for artists to experiment with the latent space, offering the ability to fine-tune the noise characteristics to achieve desired artistic effects. This node is particularly useful for generating variations of an image or exploring creative deviations from a standard sampling process.

Latent Noise Injection Sampler (CRT) Input Parameters:

enable_noise_injection

This parameter determines whether noise injection is enabled or disabled during the sampling process. When set to "enable," noise is injected into the latent space, allowing for more varied and potentially creative outputs. The default value is "disable," which means standard sampling without noise injection. This parameter is crucial for deciding whether to introduce randomness into the sampling process.

injection_point

This parameter specifies the percentage of the total sampling steps after which noise will be injected. It is a float value ranging from 0.0 to 1.0, with a default of 0.75, meaning noise will be injected after 75% of the steps. Adjusting this value allows you to control when the noise affects the latent space, which can significantly impact the final output's characteristics.

injection_seed_offset

This integer parameter adds an offset to the main seed used for generating the noise pattern. It ranges from -100 to 100, with a default value of 1. This offset allows for variations in the noise pattern, enabling you to explore different noise configurations without changing the main seed, thus providing a way to fine-tune the randomness introduced.

injection_strength

This float parameter controls the strength of the injected noise, with a range from -20.0 to 20.0 and a default value of 0.25. The strength determines how much the noise will influence the latent space, with higher values leading to more pronounced effects. Adjusting this parameter allows you to balance between subtle and strong noise influences on the output.

normalize_injected_noise

This parameter decides whether the injected noise should be normalized to match the latent's mean and standard deviation. It can be set to "enable" or "disable," with "enable" as the default. Normalizing the noise ensures that it integrates smoothly with the existing latent space, maintaining the overall statistical properties while introducing randomness.

Latent Noise Injection Sampler (CRT) Output Parameters:

final_latent

This output represents the final latent space after the noise injection and sampling process. It contains the modified latent samples that have been influenced by the injected noise, reflecting the changes made during the sampling stages. This output is crucial for understanding how the noise injection has altered the latent space and is used for further processing or decoding into an image.

decoded_image

The decoded_image output is the final image generated from the modified latent space. It represents the visual result of the entire sampling process, including the effects of noise injection. This output is essential for evaluating the artistic impact of the noise injection and serves as the final product of the node's operation.

Latent Noise Injection Sampler (CRT) Usage Tips:

  • Experiment with different injection_point values to see how the timing of noise injection affects the final image. Early injection points can lead to more dramatic changes, while later points might result in subtler variations.
  • Adjust the injection_strength to control the intensity of the noise effect. Start with the default value and gradually increase or decrease it to find the right balance for your artistic goals.

Latent Noise Injection Sampler (CRT) Common Errors and Solutions:

"Denoise value too low, using minimum 1 step."

  • Explanation: This warning occurs when the denoise value is set too low, resulting in zero actual steps for sampling.
  • Solution: Ensure that the denoise value is set high enough to allow for at least one sampling step.

"Injection point at or beyond total steps - running standard sampling..."

  • Explanation: This error happens when the injection point is set at or beyond the total number of sampling steps, making noise injection impossible.
  • Solution: Adjust the injection_point to a value less than the total number of steps to enable noise injection.

"Original std too small, skipping normalization"

  • Explanation: This warning indicates that the original standard deviation of the latent samples is too small for effective normalization of the injected noise.
  • Solution: Consider adjusting the noise parameters or ensuring that the latent space has sufficient variability before injection.

Latent Noise Injection Sampler (CRT) Related Nodes

Go back to the extension to check out more related nodes.
CRT-Nodes
RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

Latent Noise Injection Sampler (CRT)