Steaked-nodes

ComfyUI Custom Nodes

A collection of powerful image processing nodes for ComfyUI, including distortion effects, texture generation, edge detection, blending, and retro effects.

Installation

1. Clone or download this repository into your ComfyUI custom_nodes folder:

cd ComfyUI/custom_nodes
git clone https://github.com/StealthNinja1O1/Steaked-nodes.git

2. Restart ComfyUI

Or do so from the Comfy Manager -> Install via Git URL

Available Nodes

1. Load and Crop Image (LoadAndCrop)

Load images with advanced cropping capabilities.

Features:

• Load images from file
• Crop with adjustable dimensions and position

---

2. Color Blender (ColorBlender)

Advanced color manipulation and blending.

Features:

• Color overlay with multiple blend modes
• Hue/Saturation/Brightness adjustments
• Color temperature control
• Tint and vibrance adjustments

---

3. Image Distortion (ImageDistortion)

Apply 12 different distortion effects to images.

Distortion Types:

• Wave - Sinusoidal wave patterns
• Swirl - Rotational swirl effect (supports offset)
• Kaleidoscope - Mirrored kaleidoscope segments
• Pixel Sort - Sort pixels by brightness
• Displacement - Noise-based displacement mapping
• Fisheye - Fisheye lens distortion
• Ripple - Concentric ripple effect (supports offset)
• Twist - Spiral twist effect
• Spherize - Spherical warping
• Glitch - Digital glitch artifacts
• Mosaic - Mosaic tile effect
• Warp - Multi-octave noise warping

Parameters:

• intensity (0-100) - Effect strength
• frequency (0-100) - Pattern frequency/detail
• offset_x/y (-50 to 50) - Center offset for applicable effects

5. Edge Detection (EdgeDetection)

Detect edges with multiple algorithms and styling options.

Algorithms:

• Sobel - Standard gradient-based edge detection
• Prewitt - Similar to Sobel, different kernel
• Scharr - Improved Sobel with better rotational symmetry
• Roberts - Fast 2×2 gradient operator
• Laplacian - Second derivative edge detection
• Canny - Multi-stage edge detection with non-maximum suppression

Parameters:

• threshold (0-1) - Edge sensitivity
• thickness (1-5) - Edge line thickness
• invert - Swap edge/background colors
• color RGB - Edge line color
• background RGB - Background color

Use Cases:

• Line art generation
• Outline extraction
• Artistic edge effects
• Preprocessing for other effects

---

6. Image Blender (ImageBlender)

Blend two images with professional blend modes.

Blend Modes:

• Normal - Standard alpha blending
• Multiply - Darkens (multiply colors)
• Screen - Lightens (inverse multiply)
• Overlay - Contrast-preserving blend
• Soft Light - Gentle contrast adjustment
• Hard Light - Strong contrast adjustment
• Color Dodge - Brightens highlights
• Color Burn - Darkens shadows
• Darken - Keep darker pixels
• Lighten - Keep lighter pixels
• Difference - Absolute difference
• Exclusion - Similar to difference, softer
• Add - Additive blending
• Subtract - Subtractive blending

Parameters:

• base_image - Bottom layer
• blend_image - Top layer (auto-resized to match)
• blend_mode - Blend algorithm
• opacity (0-1) - Blend strength
• mask (optional) - Grayscale mask for selective blending

Use Cases:

• Composite multiple generations
• Texture overlay
• Color grading
• Creative blending effects

---

7. Halftone Effect (HalftoneEffect)

Apply retro printing and dithering effects.

Effect Types:

• Halftone Dots - Classic circular halftone print
• Halftone Lines - Parallel line pattern
• Bayer Dithering - 4×4 ordered dithering
• Floyd-Steinberg - Error diffusion dithering
• Ordered Dithering - 8×8 ordered dithering
• Newspaper - Newspaper print simulation
• Crosshatch - Multi-directional hatching

Parameters:

• dot_size (2-20) - Size of halftone elements
• angle (0-360°) - Rotation angle for directional effects
• sharpness (0.1-2.0) - Dot sharpness/contrast
• contrast (0.5-2.0) - Input contrast adjustment

Use Cases:

• Retro/vintage effects
• Print simulation
• Comic book style
• Pop art effects
• Monochrome artistic styles

8. Regional Prompts (RegionalPromptsLatent, RegionalPromptsAttention)

Define up to 4 rectangular regions on a canvas to apply different prompts to specific areas of an image. Each region has its own prompt, weight, and timestep scheduling.

Two Variants:

1. Regional Prompts (Latent) - Uses ComfyUI's standard latent masking mechanism. Each regional prompt is encoded separately and combined with masks. Better isolation between regions, but may have visible seams at boundaries.

2. Regional Prompts (Attention) - Uses attention hooks to modify the attention computation during sampling. Faster with more coherent blending, but some "bleeding" between regions may occur. Recommended for most use cases.

Features:

• Interactive Canvas: Drag boxes to move, drag edges/corners to resize
• Per-Region Controls:
  • Weight (0.0-2.0): Strength multiplier for the region's influence
  • Start/End (0.0-1.0): Timestep range when the regional prompt is active
    • Start: When the prompt begins affecting generation (0.0 = first step)
    • End: When the prompt stops affecting generation (1.0 = last step)
    • Example: Start: 0.0, End: 0.3 applies the prompt only during the first 30% of sampling
• Automatic Mask Handling: Overlapping regions are automatically normalized
• Base Prompt Fill: Unmasked areas are automatically filled with the base prompt

Inputs:

• base_prompt: Main prompt applied everywhere (or as fallback for unmasked areas)
• prompt_1 to prompt_4: Prompts for each of the 4 available boxes
• width/height: Canvas dimensions in pixels
• clip: CLIP model for text encoding

---