ComfyUI nodes for flexible noise control in image generation workflows
Top 90.0% on sourcepulse
This repository provides six custom nodes for ComfyUI, designed to offer enhanced control and flexibility over the noise generation and manipulation process. It targets users looking to implement advanced image generation techniques such as creating subtle variations of existing outputs or reconstructing the noise parameters of a given image.
How It Works
The nodes facilitate fine-grained control over latent space noise. Key nodes include Noisy Latent Image
for generating noise with specified parameters (source, seed, dimensions), Slerp Latents
for interpolating between two latent batches, and Inject Noise
for applying generated or manipulated noise to latents. The Unsampler
node performs the inverse of a sampling process, calculating the noise required to generate a target image, enabling "un-sampling" workflows. The Get Sigma
node assists in determining appropriate noise strengths for sampler integration.
Quick Start & Requirements
custom_nodes
directory.example_workflow
folder.Highlighted Details
Unsampler
node allows for the reconstruction of noise from an existing image, facilitating "un-sampling."Slerp Latents
and Get Sigma
nodes enable sophisticated noise blending and strength calculation for iterative denoising.Maintenance & Community
Licensing & Compatibility
Limitations & Caveats
add_noise
and return_with_leftover_noise
settings should be disabled.1 year ago
1 week