OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Skimmed Cfg

Skimmed Cfg is a Python collection of custom nodes for ComfyUI that alters classifier-free guidance in latent-diffusion image-generation workflows. Its nodes replace or interpolate values from positive and negative predictions, offer separate positive and negative scale controls, adjust differences by scale, or flip guidance over time. The project aims to help manage over-guidance effects while changing prompt influe

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Skimmed Cfg — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a ComfyUI latent-diffusion workflow needs alternatives to standard CFG behavior, especially when experimenting with higher guidance scales or reducing burn.

Who it is for

  • ComfyUI users working with latent-diffusion image generation.
  • Creators experimenting with CFG scales, prompt adherence, saturation, randomness, or anti-burn adjustments.

Documented uses

  • Add a guidance-adjustment node after a model loader.
  • Replace or interpolate selected negative-prediction values with positive-prediction values.
  • Tune positive and negative influence with separate scales.
  • Experiment with scale-dependent adjustments or timed-flip guidance.
Before you choose

Requirements & limits

  • The repository provides several node variants with different guidance behaviors; choose based on whether the goal is replacement, interpolation, scale adjustment, or timed-flip enhancement.
  • The README places the node after the model loader.
  • Creator suggestions vary by desired output: roughly 2–3 for stronger anti-burn, 4 as a middle setting, and 5–7 for colorful or strong styles.
  • Sampler and step count matter: SDE samplers may still burn, while low step counts can produce poor results.
  • The README specifically reports SD3.5 turbo testing and suggests a 2–2.5 skimming scale for it.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Place a Skimmed CFG node in a ComfyUI workflow

Add the node after the model loader and set an initial skimming scale for image-generation experiments.

Before you start

  • ComfyUI workflow with a model loader and latent-diffusion image-generation setup.
  • Skimmed Cfg custom nodes available in ComfyUI.

Steps

  1. Add a Skimmed CFG node to the workflow.

  2. Connect it after the model loader, as shown in the repository README.

  3. Set a skimming scale; the README suggests 2–3 for stronger anti-burn, 4 as a middle setting, or 5–7 for colorful or strong styles.

  4. Generate images and adjust the scale to suit the workflow.

Expected outcome

The workflow applies the selected Skimmed CFG node after the model loader, allowing CFG guidance to be adjusted during generation.

Guide limitations

  • The suggested scales are creator-provided tips and may not suit every model or sampler.
  • SDE samplers may still burn; low step counts can cause poor results, and low skimming scales may require more steps.
  • The README reports testing with SD3.5 turbo and suggests 2–2.5 for that model; it does not establish compatibility across all models.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
8 October 2026
Original website title
Extraltodeus/Skimmed_CFG · Official homepage title · retained in its source language
How we review sourcesSuggest a correction