Visit ComfyUI Online for ready-to-use ComfyUI environment
ComfyUI-Latent-Modifiers enhances the diffusion process in ComfyUI by providing nodes that modify latent variables, including adjustments for sharpness, tonemap, rescale, and extra noise.
ComfyUI-Latent-Modifiers is an extension designed to enhance the capabilities of ComfyUI by providing a set of nodes that modify the latent space during the diffusion process. This extension allows AI artists to have more control over the quality and style of their generated images. By adjusting various parameters, users can achieve sharper details, better contrast, and more stylistic effects without needing deep technical knowledge.
The extension works by introducing multiple parameters that influence the latent space during the diffusion process. Think of the latent space as a hidden layer where the image is gradually formed. By modifying this space, you can steer the final output towards your desired quality and style. The extension provides tools to adjust sharpness, contrast, noise, and other aspects, making it easier to fine-tune the generated images.
This feature sharpens the noise in the middle of the diffusion process, leading to more perceptual detail, especially at higher strengths. It uses code from Fooocus's sampling process.
Tonemapping clamps the conditioning noise (CFG) using a user-chosen method, allowing for higher CFG values. Different methods like Reinhard, Arctan, Quantile, Gated, CFG-Mimic, and Spatial-Norm offer various ways to control the noise.
This feature dynamically lowers the CFG by comparing the standard deviation to the existing latent, helping to maintain a balanced image quality.
Adds extra noise in the middle of the diffusion process to conditioning and performs the inverse operation on unconditioning if chosen. This can add interesting variations to the generated images.
Adjusts the contrast of the conditioning, leading to more vibrant and pop-style results. It acts as a secondary CFG slider for stylization without changing the subject's pose and location much.
As CFG increases, the mean can drift away from zero. This feature subtracts the mean or median of the latent, potentially leading to sharper and higher frequency results but may cause discoloration.
Normalizes the latent using avg_pool2d, reducing noisy artifacts and enhancing features like sharpness.
Converts the latent to frequencies, clamps higher frequencies while boosting lower ones, and then converts it back to an image latent. This helps to address oversaturation or burning due to higher CFG values.
Currently, the extension does not include different models but focuses on providing a comprehensive set of features within a single framework. Each feature can be customized to achieve various effects, making it versatile for different artistic needs.
Extra Noise
and Spectral Modulation
settings. Lower the spectral_mod_percentile
and spectral_mod_multiplier
to reduce noise.Sharpness
parameter. Ensure that Combat CFG Drift
is properly configured to avoid discoloration.Spectral Modulation
feature to clamp higher frequencies. Adjust the tonemap_multiplier
and tonemap_percentile
for better control.Contrast
settings. Use positive values for more contrast and negative values for less.Can I use multiple features simultaneously? Yes, the extension is designed to allow multiple features to work together seamlessly.
What are the recommended settings for beginners?
Start with default values and gradually adjust parameters like Sharpness
, Contrast
, and Tonemap
to see their effects.
For additional resources, tutorials, and community support, you can explore the following:
© Copyright 2024 RunComfy. All Rights Reserved.