395ecc93dac0eb45f43c01cb56ed203d60764598
Added affect_uncond (for sharpness) Made extra noise affect uncond Added low-pass for extra_noise (May go away?) Added combat cfg drift (mean centering) Added a new contrast method (use negative values for usage) Added divisive normalization, which can help with extreme added values (like sharpness)
ComfyUI-Latent-Modifiers
A repository of ComfyUI nodes which modify the latent during the diffusion process.
Latent Diffusion Mega Modifier (sampler_mega_modifier.py)
Adds multiple parameters to control the diffusion process towards a quality the user expects.
-
Sharpness: utilizes code from Fooocus's sampling process to sharpen the noise in the middle of the diffusion process. This can lead to more perceptual detail, especially at higher strengths.
-
Tonemap: Clamps conditioning noise (CFG) using a user-chosen method, which can allow for the use of higher CFG values.
-
Rescale: Scales the CFG by comparing the standard deviation to the existing latent to dynamically lower the CFG.
-
Extra Noise: Adds extra noise in the middle of the diffusion process, akin to how sharpness sharpens the noise.
Languages
Python
100%