OpenViSTAI · Agents · Blockchain · Crypto
Menu
AI Developer Tools & SDKs

Comfyui Automaticcfg

Comfyui Automaticcfg is a Python-based set of custom nodes for ComfyUI that changes classifier-free guidance behavior during Stable Diffusion image sampling. Its self-rescaling CFG aims to steer output intensity, while additional nodes provide negative-strength interpolation, boost behavior, and advanced controls.

Independent OpenViST project profile, based on linked sources.

Comfyui Automaticcfg — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it in a ComfyUI image-generation workflow when you want to test alternative CFG behavior, reduce burn-like artifacts, or experiment with the extension’s conditional guidance and speed controls.

Who it is for

  • ComfyUI users who want to experiment with CFG rescaling and image intensity.
  • Stable Diffusion workflow builders exploring guidance-related sampling speed options.

Documented uses

  • Rescale CFG across sampling steps to influence overall image intensity.
  • Use negative-strength interpolation to adjust the relationship between conditioned and unconditioned guidance.
  • Experiment with a boost option that disables negative guidance near sigma values of 1.
Before you choose

Requirements & limits

  • The extension changes the meaning of the sampler’s CFG scale; the README suggests 8 as a starting point, not a universal setting.
  • The project reports PixArt Sigma support.
  • For SD3, the README specifies use without boost or unconditional-guidance cutting.
  • Presets are evaluated with eval() and may execute malicious code, so the project advises trusting the preset source.

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

Use AutomaticCFG in a ComfyUI workflow

Connect AutomaticCFG to a sampler and try its self-rescaling CFG behavior.

Before you start

  • A ComfyUI workflow with a sampler.

Steps

  1. Add the AutomaticCFG node from the ComfyUI model_patches category to the workflow.

  2. Connect the node to the sampler so it can rescale CFG during sampling.

  3. Try setting the sampler CFG to 8 as the README’s starting point; adjust it to explore changes in output intensity, brightness, saturation, or sharpness.

Expected outcome

The connected node changes the sampler’s CFG behavior through per-step rescaling, with CFG serving as a control for output intensity rather than its usual meaning.

Guide limitations

  • CFG values no longer behave as ordinary CFG values; results may differ from standard sampling.
  • The README describes CFG 8 as a suggested starting point, not a guaranteed result.
  • For SD3, do not use boost or cut the unconditional guidance; the README says 20 steps works nicely.
  • Presets are interpreted with eval() and may execute malicious code, so only use presets from trusted sources.
Source-based practical guide

Enable boost guidance behavior

Use the boost toggle to disable negative guidance near sigma values of 1.

Before you start

  • An AutomaticCFG node connected in a ComfyUI workflow.

Steps

  1. Enable the node’s boost toggle.

  2. Run the workflow with the node connected; boost turns off negative guidance when sigmas are near 1.

Expected outcome

Negative guidance is disabled near sigma 1; the README says this doubles inference speed.

Guide limitations

  • The speed claim is the repository author’s reported behavior and is not a guaranteed result.
  • Do not use boost with SD3, according to the README’s stated limitation.
Evidence and transparency

Sources & review

Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
Extraltodeus/ComfyUI-AutomaticCFG · Official homepage title · retained in its source language

Related options

Skimmed_CFG — The repository README recommends moving to this project.

pre_cfg_comfy_nodes_for_ComfyUI — The README says most other functionalities were recreated in this repository.