virtuoso-nodes

Virtuoso Nodes for ComfyUI

This set of nodes is designed to give some Photoshop-like functionality within ComfyUI. The nodes available are:

• Blend Modes: Applies an image to another image using a blend mode operation. Every conceivable blend mode is available.
• Blend If: Composites one image on top of another with transparency based on several parameters.
• Adjustment Nodes
  • Selective Color: Adjust the color of a specific color or brightness range in an image, as with Photoshop’s Selective Color adjustment layer.
  • Color Balance: Make detailed color balance adjustments to shadows, midtones, and highlights of an image.
  • Color Balance Advanced: Color balance of a targeted brightness range.
  • Levels: Adjust the brightness levels of an image or single color channels. Works the same as Photoshop’s Levels adjustment layer.
  • Black and White: Transform a color image into Black and White while controlling brightness levels based on hue. Works the same as Photoshop’s Black and White adjustment layer.
  • Hue/Saturation: Simplified version of the Advanced Hue/Saturation Node. Allows you to choose colors by name, and choose from preset range sizes and feather values.
  • Hue/Saturation Advanced: Control Hue, Saturation, and Lightness of an image based on the selection of a range of hues. Works the same as Photoshop’s Hue/Saturation adjustment layer.
• Blur Nodes
  • Motion Blur / Motion Blur with Depth Map
  • Lens Blur / Lens Blur with Depth Map
  • Gaussian Blur / Gaussian Blur with Depth Map
• Solid Color Nodes
  • Solid Color: Create a solid color image by choosing from a list of 16 basic colors.
  • Solid Color RGB: Create a solid color image by entering Red, Green, and Blue values, or entering an RGB hex value.
  • Solid Color HSV: Create a solid color image by entering Hue, Saturation, and Value.
• Channel Nodes
  • SplitRGB: This node takes an image and splits it into its red, green, and blue components. These then can be used in creative ways with blend modes as well.
  • MergeRGB: This node takes three images, and merges the specified channels into one image. It is the complementary node to the SplitRGB node, to be used to recombine channels that you have split, though you can feel free to merge any images you wish for trippy effects.

Installation Instructions

Blend Modes

This node gives access to thirty different blend modes, and includes a scaling option for the top layer and the ability to use external masks.

Node controls:

• backdrop: This is the background image. The output image will have the same dimensions as this image.
• source: This is the top image. If it is not the same size as the backdrop image, it will be resized to cover the backdrop.
• mask: Can be used to retain the alpha channel from the source image, or you can use an external mask to control the blended area.
• opacity: The opacity of the source image. The backdrop image is always 100% opacity.
• source_adjust: The method by which the source image will be resized if necessary.
  • Stretch will change each dimension independently just as much as needed to match that dimension, potentially changing the aspect ratio of the source image and causing distortion.
  • Crop will maintain the aspect ratio of the source image, and resize it until it just covers the backdrop image, then crop what doesn’t fit.
• blend_mode: normal is the default blend mode.

Examples

Normal Darken Lighten Contrast Component Specialty

Learn about the math behind blend modes
A practical discussion of the blend modes

Blend If Node

This node gives you the functionality that is available with Photoshop’s Blend If feature:

This feature composites the top_layer image over the bottom_layer image based on various possible calculations. ComfyUI’s interface options being limited, the inputs have all been converted to fields instead of nice sliders.

Node input controls:

• top_layer: This is the top image. If it is not the same size as the bottom_layer image, it will be resized to cover the bottom layer. This is equivalent to ‘current layer’ in Photoshop.
• bottom_layer: This is the background image. The output image will have the same dimensions as this image. This is equivalent to ‘underlying layer’ in Photoshop.
• mask: This is an optional mask for the top layer, which will add to the transparency of the top layer in addition to the transparency created by the Blend If process. Any alpha channels in the top_layer and bottom_layer images are ignored.
• blend_if_layer: Choose the layer that will be analyzed to determine the opacity of the top layer. This is equivalent to Photoshop’s ‘current layer’ and ‘underlying layer’ sliders. This node only calculates one or the other.
• blend_if_channel: Which color channel (or gray) will be used to analyze the selected layer. This is equivalent to the “Blend If:” dropdown in photoshop.
• start_rise, end_rise, start_fall, end_fall: These four inputs define which brightness values for the selected channel on the selected layer are opaque. They are equivalent to the triangles below the ‘current layer’ and ‘underlying layer’ sliders, which each split into two to add softness to the selection. In this node, any brightness values lower than start_rise or greater than end_fall will be completely transparent, and brightness values between end_rise and start_fall will be completely opaque. Brightness values between start_rise and end_rise and between start_fall and end_fall will be partially transparent to feather the edges of the selection.
• opacity: Values less than 1 will reduce the opacity of the top_layer image in addition to any transparency calculated by the Blend If process. This is the equivalent to the opacity setting for the entire layer in Photoshop.
• match_size: Method to use to resize the top_layer image if it is not the same size as the bottom_layer image.
• invert_mask: ‘yes’ will invert the incoming mask before applying it to the top_layer image. Set to ‘yes’ by default, as it seems to be the more common situation.

Node Outputs

• IMAGE: The final image with top_layer composited on top of bottom_layer.
• MASK: This is the mask generated by the Blend If process. It can be useful to ensure you are getting the result you want, or to do further processing on the same pixels.
  

Adjustment Nodes

These nodes replicate Photoshop adjustment layers.

Selective Color Node

This feature allows you to adjust the color balance of a specified range of colors or brightness in an image. This node can be chained to affect different color ranges.

Node input controls:

• image: This is image to be adjusted.
• color_range: This selects which range of colors or gray values will be affected.
• cyan, magenta, yellow, black: These inputs control how much of each secondary color or black is added or removed from the color range selected.
• method: ‘absolute’ will directly apply the change, whereas ‘relative’ will apply the change as a percentage of the current value, resulting in a subtler effect.

Good reference on selective color

Color Balance Node

This feature allows you to finely adjust the color balance of the shadows, midtones, and highlights of an image. It reproduces the Photoshop functionality of the Color Balance layer.

Node input controls:

• image: This is image to be adjusted.
• lows, mids, highs: Each of these prefixes indicates the brightness range on which the correction will have the greatest effect.
• cyan_red, magenta_green, yellow_blue: These suffixes indicate the color axis the adjustment will work on. -1 will be the most cyan/magenta/yellow possible, and 1 will be the most red/green/blue possible.
• preserve luminosity: This will maintain the brightness of the image while adjusting the relative color values. This will help prevent clipping of brightness values, while making the overall effect more subtle.
  

Color Balance Advanced Node

This node allows you to target a specific brightness as the center of an adjustment range, and gives full range of adjustment. Unlike the standard Color Balance Node, it is easy to create distorted corrections with this node if you are not careful.

Node input controls:

• image: This is image to be adjusted.
• brightness_target: What brightness level the correction should be centered on. 0 is black and 1 is white. This brightness level will receive the strongest correction, tapering off for brighter and darker pixels.
• cyan_red, magenta_green, yellow_blue: These suffixes indicate the color axis the adjustment will work on. -1 will be the most cyan/magenta/yellow possible, and 1 will be the most red/green/blue possible. Unlike the standard Color Balance Mode, these adjustments are unrestricted, so less is more!
• preserve luminosity: This will maintain the brightness of the image while adjusting the relative color values. This will help prevent clipping of brightness values, while making the overall effect more subtle.
  

Levels

This node works just like Photoshop’s Levels, without the fancy controls. Move around your black and white points and your gamma, across all channels, or select just one color channel.

Node controls:

• image: The image to set the levels for.

• channel: Choose ‘RGB’ to work on the whole image, or a color to set levels only on that color channel.

• input_black_point: Choose what brightness level becomes black. 0 is black, 1 is white.

• input_gamma: Adjust the gamma.

• input_white_point: Choose what brightness level becomes white. 0 is black, 1 is white.

• output_black_point: After input calculations are complete, what brightness will black pixels be output at.

• output_white_point: After input calculations are complete, what brightness will white pixels be output at.

  
  

Black and White

This node works just like Photoshop’s Black and White adjustment layer. Add or remove brightness based on each of the color ranges in the image.

Node controls:

• image: The image to set the levels for.

• red, green, blue, cyan, magenta, yellow: Reduce these values to make the given color range darker, increase the values to make the given color range brighter.

  
  

Hue / Saturation

This node works like a simplified version of Photoshop’s Hue/Saturation adjustment layer. Select a color range to work on, then change the hue, saturation, and lightness of that range. Note that Stable Diffusion images seem to have pretty brutal color compression, so this node is best for subtle effects… otherwise things get pretty weird.

Node input controls:

• image: The image to set the levels for.
• hue: Choose the main color to affect.
• hue_width: How far away from the main color do you want to full affect?
• feather: How much softening do you want from fully selected colors to unselected colors?
• hue_offset: This value changes each pixel’s hue value by rotating it this number of degrees around the color wheel. Each color on the hue list is 60 degrees apart.
• sat_offset: This value increases or decreases color saturation.
• lightness_offset: This value affects how much white or black is added to the image.

Node Outputs

• IMAGE: The final image
• MASK: This is the mask generated by the hue selection process. It can be useful to ensure you are getting the result you want, or to do further processing on the same pixels.

Hue / Saturation Advanced

This node works like Photoshop’s Hue/Saturation adjustment layer. Select a color range to work on, then change the hue, saturation, and lightness of that range. Note that Stable Diffusion images seem to have pretty brutal color compression, so this node is best for subtle effects… otherwise things get pretty weird.

To use this node well, you need to be aware of where the colors are placed on the color wheel. The measurements are in degrees:

• Red: 0 degrees (and 360 degrees)
• Yellow: 60 degrees
• Green: 120 degrees
• Cyan: 180 degrees
• Blue: 240 degrees
• Magenta: 300 degrees

Node input controls:

• image: The image to set the levels for.
• hue_low: What hue (in degrees) will start the range of fully selected hues?
• hue_low_feather: How many hue degrees will the low end of the selected hue range take to feather from fully selected to unselected?
• hue_high: What hue (in degrees) will end the range of fully selected hues? Note that because the hue wheel is a circle, the hue_high value can be lower than the hue_low value.
• hue_high_feather: How many hue degrees will the high end of the selected hue range take to feather from fully selected to unselected?
• hue_offset: This value changes each pixel’s hue value by rotating it this number of degrees around the color wheel.
• sat_offset: This value increases or decreases color saturation.
• lightness_offset: This value affects how much white or black is added to the image.

Node Outputs

• IMAGE: The final image
• MASK: This is the mask generated by the hue selection process. It can be useful to ensure you are getting the result you want, or to do further processing on the same pixels.

Blur Nodes

Nodes to blur the image in various ways.

Motion Blur / Motion Blur with Depth Map

These two nodes apply motion blur to the image. The Motion Blur with Depth Map takes a depth map image which is used to apply the blur selectively. The darker areas in the depth map will be blurred less, and the lighter areas will be blurred more.

Node input controls:

• size: The strength of the blur (Motion Blur node only)
• angle: Angle of motion for the blur
• num_layers: How many levels of depth are calculated for the blur. More layers creates a smoother blur, but can add a lot of time. (Motion Blur with Depth Map node only)
• min_blur: The blur percentage of the darkest area of the depth map. (Motion Blur with Depth Map node only)
• max_blur: The blur percentage of the lightest area of the depth map. (Motion Blur with Depth Map node only)
  

Lens Blur / Lens Blur with Depth Map

These two nodes apply lens blur to the image. The Lens Blur with Depth Map takes a depth map image which is used to apply the blur selectively. The darker areas in the depth map will be blurred less, and the lighter areas will be blurred more.

Node input controls:

• radius: The strength of the blur (Lens Blur node only)
• components: The number of lens components in the virtual lens
• exposure_gamma: Effects the quality of the blurred areas
• num_layers: How many levels of depth are calculated for the blur. More layers creates a smoother blur, but can add a lot of time. (Lens Blur with Depth Map node only)
• min_blur: The blur percentage of the darkest area of the depth map. (Lens Blur with Depth Map node only)
• max_blur: The blur percentage of the lightest area of the depth map. (Lens Blur with Depth Map node only)
  

Gaussian Blur / Gaussian Blur with Depth Map

These two nodes apply lens blur to the image. The Lens Blur with Depth Map takes a depth map image which is used to apply the blur selectively. The darker areas in the depth map will be blurred less, and the lighter areas will be blurred more.

Node input controls:

• amount: The strength of the blur (Gaussian Blur node only)
• sigma: The strength of the blur (Gaussian Blur with Depth Map node only)
• num_layers: How many levels of depth are calculated for the blur. More layers creates a smoother blur, but can add a lot of time. (Gaussian Blur with Depth Map node only)
• min_blur: The blur percentage of the darkest area of the depth map. (Gaussian Blur with Depth Map node only)
• max_blur: The blur percentage of the lightest area of the depth map. (Gaussian Blur with Depth Map node only)

Solid Color Nodes

Creates an image of one color.

Solid Color

This node is a simple way to create a solid color image by choosing a color from a list.

Node input controls:

• color: The color to make the new solid color image.
• height: Pixel height of new solid color image.
• width: Pixel width of new solid color image.

Solid Color RGB

This node creatse a solid color image by setting the Red, Green, and Blue values, or one hexidecimal value.

Node input controls:

• red, green, blue: The amount of each primary color to include in the color image. Values are 0-255, with 0 meaning none of the primary color and 255 being 100% of that color.
• height: Pixel height of new solid color image.
• width: Pixel width of new solid color image.
• hex: A hexidecimal code containing an RGB value. If this value is provided and is valid, it will be used instead of the above red, green, and blue values.

Solid Color HSV

This node creatse a solid color image by setting the hue, saturation, and value of the color. To use this node well, you need to be aware of where the colors are placed on the color wheel. The measurements are in degrees:

• Red: 0 degrees (and 360 degrees)
• Yellow: 60 degrees
• Green: 120 degrees
• Cyan: 180 degrees
• Blue: 240 degrees
• Magenta: 300 degrees

Note that any color with the value set to 0 will be black, but not every color with the value set to 100 will be white. If you want white, use value = 100 and saturation = 0.

Node input controls:

• hue: The degree value of the hue you want.
• saturation: The percent of color saturation you want for the color.
• value: The brightness of the color, in percentage from 0-100.
• height: Pixel height of new solid color image.
• width: Pixel width of new solid color image.

Channel Nodes

Nodes that work on the color channels of the image.

Split RGB and Merge RGB

These nodes split apart images into their component color channels and merge them back together. You can do fun things with the layers once they’re separate, and then combine them back together when you’re done. If you want even more control when merging them back together, you can use the Linear Dodge (Add) blend mode instead (twice), which will accomplish the same thing while giving you the ability to play around with the opacities of each layer.

Node controls:

• image: The image to split on the Split RGB node, and the merged image on the Merge RGB node
• red, green, blue: The separate images holding the red, green, and blue channels of the image. Note that there is nothing stopping you from using any images you want here on the Merge RGB node-- the node will only take the indicated color channel from each image.

Installation:

1. Git clone this repo into a folder in ComfyUI\custom_nodes.
2. pip install -r requirements.txt.
3. Install FFmpeg, which is required for the Selective Color Node.

Please let me know if you have any thoughts or suggestions!