Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > ComfyUI-DeGrid > VAE DeGrid (Nyquist Notch)

ComfyUI Node: VAE DeGrid (Nyquist Notch)

Class Name

VAEDeGrid

Category
image/postprocessing
Author
lunaaispace-eng (Account age: 379days)
Extension
ComfyUI-DeGrid
Latest Updated
2026-10-28
Github Stars
0.1K

How to Install ComfyUI-DeGrid

Install this extension via the ComfyUI Manager by searching for ComfyUI-DeGrid
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter ComfyUI-DeGrid in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

VAE DeGrid (Nyquist Notch) Description

Removes two-pixel VAE grid artifacts after decoding while preserving edges and skipping clean images.

VAE DeGrid (Nyquist Notch):

VAEDeGrid reduces the fine, repeating two-pixel grid that can appear after decoding images with Qwen Image, Qwen Image 2.1, or Wan 2.1 VAEs. It is designed as a postprocessing step immediately after VAE Decode, before resizing, sharpening, upscaling, or restoration makes the lattice harder to address or treats it as image detail. Its Nyquist-notch approach targets the repeating grid while protecting image edges. In its recommended automatic mode, it measures each image’s grid strength and calibrates the correction; when no grid is detected, it can leave the image untouched. This helps you remove a VAE artifact without routinely filtering images that are already clean.

VAE DeGrid (Nyquist Notch) Input Parameters:

image

The IMAGE to inspect and, when appropriate, degrid. Connect the image directly from VAE Decode for the intended workflow: the node targets a two-pixel lattice associated with the decoded image, and the provided guidance recommends running it before any resize, sharpening, or upscale. This required input has no numeric range or default.

enabled

Controls whether processing runs. The default is True. Set it to False to bypass degrid completely: the image passes through untouched, making it easy to compare the original and processed results. When bypassed, the removed-grid output is neutral gray and the status reads bypassed (enabled = off).

mode

Selects auto or manual; the default is auto. In auto, the node measures the grid in each image and sets the removal limit itself, so this is the recommended starting point. In manual, it uses the limit value you choose instead. Try manual mode only if the automatic result visibly leaves grid behind or removes more fine detail than you want.

limit

Sets the maximum per-pixel correction in manual mode on a 0–1 scale; it is ignored in auto mode. The minimum is 0.0, the maximum is 0.10, the step is 0.001, and the default is 0.02. The node’s guidance describes typical VAE-grid strength as approximately 0.005–0.02, making the default a starting point. A limit that is too low can leave grid visible in high-contrast areas; one that is too high can slightly soften fine two- to three-pixel details such as pores or fabric.

skip_when_clean

Determines whether an image with no detected grid is left completely untouched. The default is True, which is useful when a batch includes clean images: the node checks for the repeating lattice and passes clean images through without filtering. Set it to False only if you deliberately want the filter to run even when the grid detector reports no grid. This setting is separate from enabled: turning the node off always bypasses processing.

grid_gain

Controls brightness amplification of the removed_grid visualization only; it does not change the cleaned image. The minimum is 1.0, the maximum is 50.0, the step is 1.0, and the default is 10.0. Raise it when the preview appears almost flat gray and you want to inspect the faint pattern more easily. Do not use it to judge how strongly the image itself was corrected, because it affects only the preview.

grid_view

Sets the framing of the removed_grid preview. The options are full frame, 4x zoom, and 8x zoom; the default is 4x zoom. The grid is only two pixels wide, so at ordinary node-preview size, full frame can look like faint gray noise even when processing is working. The zoom options show a magnified center crop to make the lattice easier to see. This setting never crops the cleaned image.

VAE DeGrid (Nyquist Notch) Output Parameters:

image

The first IMAGE output is the cleaned image, at the same size as the input. When processing finds a grid, this is the result to send to later sharpening, upscaling, restoration, or saving steps. If enabled is False, it is the untouched input; with the default skip_when_clean setting, an image detected as clean is also passed through untouched. The node’s status text provides context about the measured grid and whether processing was skipped.

removed_grid

The second IMAGE output is a preview of what was subtracted, amplified by grid_gain and centered on gray so that a subtle correction can be inspected. A faint, uniform fine-grid or noise-like appearance is expected; in full frame view, the two-pixel pattern may be difficult to distinguish at preview size. Use 4x zoom or 8x zoom to inspect its structure. If you can recognize substantial image content such as faces or fabric in this preview, the node’s guidance suggests that a manual limit may be too high. This output is for inspection, not for further image processing; when the node is disabled, it is neutral gray.

VAE DeGrid (Nyquist Notch) Usage Tips:

  • Connect VAEDeGrid directly after VAE Decode, before resizing, sharpening, upscaling, or a restorer such as SeedVR2. This gives it the best opportunity to address the original two-pixel lattice before later steps treat it as detail.
  • Start with enabled=True, mode=auto, and skip_when_clean=True. Check the status text after a run: it reports the measured grid strength, the dominant checker or stripe orientation when a grid is detected, and whether a clean image was passed through.
  • Toggle enabled off and on to compare the result against the untouched image. If automatic mode visibly under- or over-corrects, switch to manual and adjust limit from its 0.02 default in small 0.001 steps.
  • Keep grid_view=4x zoom or choose 8x zoom when inspecting the removed pattern. Raise grid_gain if that preview is too faint; neither setting changes or crops the cleaned image.
  • In manual mode, inspect both the cleaned image and removed_grid. Grid left in the cleaned image can indicate a limit that is too low; recognizable faces or fabric in the removed-pattern preview can indicate one that is too high.

VAE DeGrid (Nyquist Notch) Common Errors and Solutions:

bypassed (enabled = off)

  • Explanation: This is a status message, not an error. Processing was disabled, so the image passed through untouched and the removed-grid preview is neutral gray.
  • Solution: Set enabled to True to run degrid, or leave it off when making an A/B comparison.

grid …/255 — none detected, passed through untouched

  • Explanation: This is a status message, not an error. The measured grid was negligible, and skip_when_clean=True prevented unnecessary filtering.
  • Solution: Leave the setting as it is if the image looks clean. If you intentionally want to run the filter regardless of detection, set skip_when_clean=False and inspect the result carefully.

grid …/255 — none detected, filtered anyway

  • Explanation: This is a status message, not an error. The measured grid was negligible, but filtering still ran because clean-image skipping was turned off.
  • Solution: Set skip_when_clean=True if you want images reported as clean to pass through untouched.

A two-pixel grid is still visible in the cleaned image

  • Explanation: The correction may be insufficient, particularly if you are using a low manual limit. Resizing or upscaling before this node also goes against the placement recommended for the original VAE lattice.
  • Solution: Place the node directly after VAE Decode and try mode=auto first. If auto visibly under-corrects, use mode=manual and increase limit gradually while checking the cleaned image.

Fine texture looks softened, or faces or fabric appear in removed_grid

  • Explanation: The filter may be removing more than the intended lattice. The node’s guidance specifically warns that a high manual limit can affect fine two- to three-pixel texture.
  • Solution: Return to mode=auto or lower the manual limit in small steps. Compare with enabled=False to check whether useful detail is being lost.

removed_grid looks like flat gray or faint noise

  • Explanation: This is not necessarily a failure. The removed artifact is very small, and its full-frame preview can be hard to read; a clean or bypassed image can also produce an uninformative preview.
  • Solution: Check the status text first. For a processed image, use 4x zoom or 8x zoom and increase grid_gain if needed. Remember that preview settings do not affect the cleaned image.

Changing limit, grid_gain, or grid_view does not change the cleaned image as expected

  • Explanation: limit is ignored in auto mode, while grid_gain and grid_view affect only the removed-grid preview. If enabled=False, processing is bypassed entirely.
  • Solution: Use mode=manual before adjusting limit for the cleaned image, and ensure enabled=True. Use grid_gain and grid_view only to make the preview easier to inspect.

VAE DeGrid (Nyquist Notch) Related Nodes

Go back to the extension to check out more related nodes.
ComfyUI-DeGrid
RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

VAE DeGrid (Nyquist Notch)