ComfyUI-MachinePaintingNodes Introduction
ComfyUI-MachinePaintingNodes is a collection of 41 nodes for editing images and organizing creative workflows in ComfyUI. It brings familiar image-editing tasks—curves, color grading, background removal, masks, blur, grain, and compositing—into your node graph. You can use it to refine an AI-generated image without moving every adjustment into a separate editor.
For example, you might remove a character’s background, soften the mask edge, place the character on a new canvas, match its colors to the scene, and add film grain. The suite also includes controls for trying different values across workflow runs and viewing image information such as brightness and color distribution.
Overview of MachinePaintingNodes
How ComfyUI-MachinePaintingNodes Works
Each node performs one job and passes its result to the next node. Think of a workflow as a stack of editing steps: an image enters, each connected node changes or inspects it, and the final result comes out at the end. You can change a setting in one step without rebuilding the entire edit.
A few ideas are especially useful when working with this suite:
- Masks control where an edit appears. A mask is like a stencil: it lets you blur, recolor, or composite selected parts of an image while leaving other areas alone. Mask tools can soften edges, expand or shrink a selection, and combine several selections.
- Blending controls how results mix. An adjustment can be mixed with the original image rather than applied at full strength. Blend modes offer different ways to combine images, colors, or grain.
- Frequency separation divides an image by scale of detail. The low-frequency layer carries broad color and tone; the high-frequency layer carries fine texture. Editing them separately can help you smooth uneven tones while preserving detail, then recombine them.
- Preview and analysis nodes help you judge an edit. In-node previews show selected effects, while a histogram and color wheel provide another way to inspect an image before making further changes.
ComfyUI-MachinePaintingNodes Features
The nodes are grouped by the kind of work they help you do.
Color adjustment and grading
Use these nodes to correct an image, establish a mood, or make images from different sources look as though they belong together.
| Node | What it does and when you might use it |
|---|---|
| Curves Adjust Pro | Adjusts brightness and color with an interactive curve editor. Edit RGB together or the red, green, and blue channels separately. It includes 17 presets, such as S-Curve Contrast, Fade, Cross Process, and Cinematic, plus mask support when the effect should appear only in part of an image. Click to add a point, drag to move it, and Shift-click to remove it. |
| Levels Adjust | Sets black and white points, midtone brightness through gamma, and output levels. Use it when an image looks washed out, too dark, or short on tonal range. |
| Auto Levels | Applies automatic levels correction as a quick starting point for tonal adjustment. |
| Selective Color Pro | Changes the color balance of specific ranges—reds, yellows, greens, cyans, blues, magentas, whites, neutrals, or blacks—using CMYK sliders. For instance, you can adjust a red garment without making the same change to every color in the image. It supports masks and has a Reset All button. |
| Brightness Contrast Adjust | Provides straightforward sliders for overall brightness and contrast. |
| Color Match Blend | Matches one image’s color to another using statistical, histogram, or Reinhard matching. Separate luminance and color controls, saturation adjustment, and blend modes let you tune how closely the images match. |
| Color Adjust Blend | Offers RGB color balance for shadows, midtones, and highlights. You can use it on its own or connect a reference image for LAB color matching, then choose a blend mode for the result. |
| LUT Apply | Applies .cube or .3dl lookup tables (LUTs), which are saved color looks. An intensity slider lets you make the look subtle or strong. Five bundled looks are listed below. |
Filters and retouching
| Node | What it does and when you might use it |
|---|---|
| Blur Pro | Offers five blur types in one node: Gaussian for general smoothing, Surface for smoothing while preserving edges, Box for a uniform blur, Median for removing speckled artifacts, and Motion for directional movement. Controls change to match the selected type. A mask limits the blur to chosen areas, optional mask blurring softens that selection, and strength mixes the result with the original. It also has an in-node preview and optimization for large blur radii. |
| Noise Grain Pro | Adds film-like grain with controls for particle size, sharpness, monochrome or color grain, and color saturation. Choose from overlay, soft light, hard light, linear light, multiply, and screen blending. A seed lets you reproduce a grain pattern, while a mask confines it to selected areas. |
| Frequency Separate | Takes an original image and a blurred version, then separates fine texture from broad color and tone. It outputs the high-frequency texture layer, the low-frequency layer, and the original image for easy chaining. You can supply the blurred image from Blur Pro or another blur node. |
| Frequency Combine | Rejoins the high- and low-frequency layers with Linear Light blending. Its opacity control sets the texture contribution: 0 leaves the smooth low-frequency result; 1 gives full reconstruction of the supplied layers. |
A simple retouching setup is to blur an image, feed the original and blurred versions into Frequency Separate, make your intended layer edits, and use Frequency Combine to bring them back together. Gaussian blur is a standard starting point; Surface Blur can be useful when you want to retain more distinct edges.
Masks, inpainting, and backgrounds
| Node | What it does and when you might use it |
|---|---|
| Remove Background Pro | Uses an AI background-removal model to separate a subject from its background. Refine the mask with grow/shrink, blur, threshold, and edge-feathering controls. Preview the masked result, mask alone, side-by-side view, or overlay. Optional rotation, zoom, and XY placement affect the image and its mask together after background removal, so the model works on the upright source. |
| Mask Editor | Refines a mask with grow, shrink, blur, and fill tools. |
| Apply Mask | Composites an image using a mask and background options. |
| Channel Mask Pro | Creates masks from an image’s red, green, blue, or alpha channel. Black point, white point, gamma, contrast, and brightness controls help isolate the area you need. It also accepts an input mask with an invert option and provides a black-and-white preview. |
| Inpaint Mask Pro | Prepares an inpainting mask and can use a separate silhouette mask to trim a rough selection to cleaner boundaries. You can invert, expand or shrink, and blur the inpaint mask; the silhouette crop has its own blur and expand/contract controls. A colored overlay preview has independent display colors and opacity for each layer. Outputs include the unchanged image, processed mask, black-and-white mask image, and preview image. |
| Mask Composite Pro 2X | Combines two masks. Each mask has its own blur, grow/shrink, and opacity settings. Screen blending creates a combined selection, while an output selector and invert toggle let you inspect or reverse the result. |
| Mask Composite Pro 6X | Provides the same controls for up to six masks. It can output the full combination or an individual mask. If mask sizes differ, they are resized to match mask_1. |
Inpaint Mask Pro is particularly useful when you have an external inpaint mask that needs refining before it enters the rest of your workflow. Its silhouette mask can act like a clean cutout around a subject, keeping a loosely painted inpaint region within that boundary.
Background removal and selective color example
Blending and positioning
- Image Blend Pro mixes images using 15 blend modes, including normal, overlay, multiply, screen, soft light, and hard light. Try different modes when layering a texture, lighting effect, or second image to change how it interacts with the image underneath.
- Transform & Rotate Pro rotates, scales, and repositions an image inside a canvas. Set canvas width and height, or use 0 to match the incoming image. You can rotate freely or snap to 45-degree increments. Its zoom control is centered at 0 for the fitted image; negative values zoom out and positive values zoom in. For reference, −50 = 0.5× and +50 = 2×. Placement is measured as a percentage of the canvas: positive X moves right, and positive Y moves up. An optional mask follows the same transform and cuts out the image over a transparent, black, or white background. Outputs are
image,mask, andmask_bw, with an in-node transparency-grid preview.
Image inspection and workflow utilities
- Histogram View displays RGB and brightness distribution in the node. Use it to inspect tonal balance while adjusting an image.
- Color Wheel View displays color distribution as a vectorscope-style view.
- Image <> Mask Switch converts between ComfyUI
IMAGEandMASKdata types. Image-to-mask creates a grayscale version; mask-to-image creates a three-channel image. Both outputs remain populated for downstream connections. - Boolean, Boolean Invert, Boolean Switch Value Output, Boolean Input Value Switch, and Boolean Master Switch provide true/false controls for switching values or routing parts of a workflow. The master switch lets one control drive multiple booleans.
- Seed Lock toggles whether the current seed value stays fixed, helping you repeat a result while testing other settings.
- Text Notes adds a title and multiline comments directly to your workflow. It is for organization and has no inputs or outputs. Text String sends text to other nodes; Show Text displays and passes along text; and Show Value displays a float, integer, string, or boolean value when you need to check what a connection is carrying.
- Mega Slider Master Value sets minimum, maximum, and step values for connected Mega Sliders. Mega Slider X1, X3, X6, and X12 give you banks of 1, 3, 6, or 12 controls with individual outputs. You can choose float or integer output and set individual slider overrides through right-click → Properties.
- Dynamic Value Range changes a value on each workflow run. Choose increment, decrement, or random mode; on reaching an end of the range, it can reverse direction or jump to the other end. It provides float, integer, and string outputs for connections that need different value types.
- Z-Image Empty Latent Image offers landscape, portrait, and square resolution presets from 512×512 to 2048×2048 for Z-Image workflows. It outputs a latent, width, and height, and includes a batch-size setting.
ComfyUI-MachinePaintingNodes Models
Remove Background Pro offers eight background-removal model choices. The names indicate useful starting points, but the best choice depends on your image; compare the mask previews when edges or subject details matter.
| Model | Suggested starting point |
|---|---|
u2net | General background-removal tasks. |
u2netp | A smaller U²-Net variant to try when model size matters. |
u2net_human_seg | Images where the subject is a person. |
u2net_cloth_seg | Clothing-focused segmentation. |
silueta | An alternative background-removal model to compare against the others. |
isnet-general-use | General-purpose subjects. |
isnet-anime | Anime-style artwork. |
sam | Another segmentation option to test when refining a subject selection. |
LUT Apply also includes five bundled color looks. These are LUTs, not AI models:
| LUT | Look to try |
|---|---|
Cinematic_Teal_Orange | A teal-and-orange cinematic grade. |
Warm_Tone | A warmer overall color mood. |
Cool_Tone | A cooler overall color mood. |
Vintage_Fade | A faded, vintage-style grade. |
High_Contrast | A stronger-contrast look. |
The bundled LUTs are placed in ComfyUI/input/luts/; the node also supports your own .cube and .3dl LUT files in that location. Use its intensity slider to keep a look from overpowering your artwork.
What's New with ComfyUI-MachinePaintingNodes
The changelog records the following releases, from newest to oldest:
| Version | Changes and why they matter |
|---|---|
| v2.2.0 | Added Transform & Rotate Pro for positioning an image and optional mask together. Remove Background Pro gained matching rotation, zoom, and placement controls after background removal. Both use the same placement convention, making it easier to move a cutout between the two nodes without changing how you think about the controls. |
| v2.1.4 | Corrected registry icon metadata to remove a broken-image avatar on the ComfyUI registry. |
| v2.1.3 | Updated dependencies and image-handling compatibility, including headless OpenCV, newer Pillow support, and a NumPy version limit. Frontend import paths were also standardized. These are primarily compatibility improvements. |
| v2.1.2 | Set Frequency Separate to the stated Photoshop Apply Image separation formula and Frequency Combine to Linear Light blending with an opacity slider. Simplified Blur Pro controls. Inpaint Mask Pro no longer supports SEGS, and its inpaint mask input became required. |
| v2.1.1 | Added Mask Composite Pro 2X and 6X for combining selections. Inpaint Mask Pro gained expand/shrink and more configurable overlay previews. Blur-type-specific controls were marked as advanced for a cleaner default view. |
| v2.1.0 | Introduced Blur Pro, Noise Grain Pro, Frequency Separate, Frequency Combine, Inpaint Mask Pro, and Image <> Mask Switch, along with the Filters category. This release added much of the suite’s retouching and mask-preparation toolkit. |
| v2.0.5 | Corrected the release version information and node count. |
| v2.0.4 | Added Z-Image Empty Latent Image with preset resolutions for Z-Image workflows. |
| v2.0.3 | Added Show Value, the Mega Slider master and slider banks, and Dynamic Value Range. These make it easier to inspect values, centralize controls, and vary settings between runs. |
| v2.0.2 | Added Text Notes, Text String, and Show Text, and organized nodes into subcategories. |
| v2.0.1 | Updated node settings and fixed display and node-structure issues. |
| v2.0.0 | Major overhaul adding LUT Apply, Channel Mask Pro, Selective Color Pro, and Boolean Invert, plus five bundled LUTs. It improved curves, previews, and categories while removing deprecated nodes. |
| v1.0.0 | Initial release with color-adjustment, blending, and boolean utility nodes. |
Troubleshooting ComfyUI-MachinePaintingNodes
An effect changes the whole image when you wanted to edit only one area.
- Check whether the node has a mask input and connect the selection you intend to use.
- Inspect the mask with a preview or a mask-to-image conversion.
- If the selection is reversed, use an available invert control.
- If its edge is too hard, soften it with the node’s mask-blur or feathering controls.
An inpaint selection spills outside the subject.
- Connect the inpaint mask to Inpaint Mask Pro.
- If you have a cleaner silhouette mask, connect it as the crop mask.
- Use the colored overlay preview to see how the two selections overlap.
- Adjust expand/shrink and blur until the processed mask follows the intended boundary, then use the processed mask output downstream.
A rotated cutout no longer seems to line up with its mask.
- Connect both image and mask to Transform & Rotate Pro so the node applies the same transform to each.
- Use its
imageandmaskoutputs together downstream. - Check the in-node transparency-grid preview. If no mask is supplied, the transformed image’s extent becomes the mask.
A mask will not connect to an image input, or an image will not connect to a mask input.
Use Image <> Mask Switch to convert to the data type the next node expects. Image-to-mask produces grayscale; mask-to-image produces a three-channel image.
Frequency-separated layers do not give the expected result when recombined.
- Supply Frequency Separate with the original image and a blurred version of that same image.
- Connect its high- and low-frequency outputs to Frequency Combine in their corresponding inputs.
- Set combine opacity to 1 when you want the full supplied texture contribution; lower it when you intentionally want a smoother result.
A transformed image moves in an unexpected direction.
In Transform & Rotate Pro and the rotation controls of Remove Background Pro, positive X moves right and positive Y moves up. The vertical direction differs from some image editors. The angle control moves counter-clockwise to the left and clockwise to the right.
Frequently asked questions
Do I need every node in a workflow?
No. Choose only the nodes needed for your edit. A simple color correction might use one curves or levels node; a more involved composite might use background removal, mask refinement, transformation, and color matching.
Which blur should I start with?
Try Gaussian for general softness, Surface when you want to smooth areas while retaining edges, Median for speckled artifacts, and Motion when you want directional blur.
Can I keep the same grain while changing other settings?
Yes. Keep the Noise Grain Pro seed unchanged to reproduce its grain pattern.
Can I apply a LUT subtly?
Yes. Reduce the LUT Apply intensity to mix less of the graded look into the image.
Can I use Transform & Rotate Pro without a mask?
Yes. Without an input mask, the node uses the transformed image extent as its mask. You can still choose a transparent, black, or white background.
Learn More about ComfyUI-MachinePaintingNodes
- ComfyUI-MachinePaintingNodes repository: the project’s source, current README, and release information.
- License details: the suite is provided under the MIT License.
