ComfyUI-KarmaNodes

Advanced KSampler node and professional post-processing tools for ComfyUI with multi-cycle sampling, progressive upscaling, and cinematic enhancement capabilities.

Overview

ComfyUI-KarmaNodes provides a comprehensive suite of nodes for advanced image generation and post-processing:

• Karma KSampler Cycle: Specialized KSampler that performs multiple sampling cycles with progressive upscaling between cycles, enabling high-quality, high-resolution image generation with better detail preservation
• Karma Film Grain: Professional film grain simulation for authentic analog film texture and cinematic aesthetics
• Karma Kolors: Advanced color grading and correction tools for professional-grade color enhancement

Features

🔄 Multi-Cycle Sampling

• Progressive Upscaling: Automatically upscales images between sampling cycles
• Configurable Cycles: Support for 2-12 sampling cycles
• Dynamic Parameters: Adjust sampling parameters between cycles

🎯 Advanced Sampling Control

• Dual Model Support: Switch between primary and secondary models at specified cycles
• Additive Conditioning: Blend additional positive/negative conditioning with strength scaling
• Dynamic Denoise: Configurable denoise strength scaling with minimum threshold protection
• Steps Scaling: Flexible steps adjustment with increment/decrement modes and threshold control
• Threshold Management: Auto-calculated or manual threshold settings for precise control

🔧 Upscaling Options

• Basic Upscaling: High-quality image rescaling with multiple resampling methods
• Model-Based Upscaling: Use dedicated upscale models for enhanced quality
• Gradual Upscaling: Multi-step upscaling for smoother transitions
• Configurable Resampling: Separate methods for image and latent space upscaling
• VAE Compatibility: Ensures dimensions are VAE-compatible (divisible by 8)

✨ Post-Processing

• Sharpening Filter: Optional unsharp mask filter between cycles
• Film Grain Effects: Realistic analog film grain simulation
• Color Grading: Professional color correction and enhancement
• Tiled VAE Support: Handle large images with tiled VAE processing
• Memory Management: Automatic device management for optimal performance

Installation

Method 1: ComfyUI Manager (Recommended)

1. Install ComfyUI Manager
2. Search for "KarmaNodes" in the manager
3. Install and restart ComfyUI

Method 2: Manual Installation

1. Navigate to your ComfyUI custom nodes directory:

   cd ComfyUI/custom_nodes/
   

2. Clone this repository:

   git clone https://github.com/karmaswint/ComfyUI-KarmaNodes.git
   

3. Install dependencies:

   cd ComfyUI-KarmaNodes
   pip install -r requirements.txt
   

4. Restart ComfyUI

Usage

Basic Workflow

1. Add the Node: Search for "Karma KSampler Cycle" in the node browser
2. Connect Inputs:
   • Connect your model, VAE, and conditioning inputs
   • Provide an initial latent image
3. Configure Parameters:
   • Set the number of cycles (2-12)
   • Configure upscale factor and method
   • Adjust denoise strengths
4. Run: Execute the workflow

Key Parameters

Core Sampling

• Steps: Number of sampling steps per cycle
• Total Cycles: Number of upscaling/sampling cycles (2-12)
• Starting Denoise: Denoise strength for the first cycle (0.0-1.0)
• Cycle Denoise: Denoise strength for subsequent cycles (0.0-1.0)

Denoise Scaling

• Enable Denoise Scaling: Automatically adjust denoise strength between cycles
• Denoise Min Threshold: Minimum denoise value to prevent going too low (0.01-1.0)

Steps Scaling

• Enable Steps Scaling: Automatically adjust sampling steps between cycles
• Steps Scaling Value: Amount to adjust steps by each cycle (1-50)
• Steps Adjustment Mode:
  • decrement: Reduce steps each cycle (for refinement)
  • increment: Increase steps each cycle (for more detail)
• Steps Threshold Mode:
  • auto: Automatically calculate threshold based on initial steps
  • manual: Use a fixed threshold value
• Steps Manual Threshold: Fixed threshold when using manual mode (1-200)

Upscaling

• Upscale Factor: Total upscaling factor to achieve
• Upscale Method:
  • basic: High-quality image rescaling
  • model: Use dedicated upscale model (requires upscale_model input)
• Scale Sampling: Resampling method for image upscaling (bilinear, bicubic, nearest, lanczos)
• Latent Upscale Method: Resampling method for latent space upscaling (bilinear, bicubic, nearest, lanczos)
• Enable Gradual Upscaling: Use multiple intermediate upscaling steps for smoother transitions
• Gradual Upscale Steps: Number of intermediate steps when gradual upscaling is enabled (1-10)

Advanced Features

• Secondary Model: Optional model to switch to at specified cycle
• Secondary Model Start Cycle: Cycle number to switch to secondary model (1-12)
• Additive Conditioning: Additional positive/negative prompts with strength control
• Sharpening: Apply unsharp mask filter between cycles

Example Workflow

[Model] → [Karma KSampler Cycle] → [VAE Decode] → [Save Image]
[VAE] ↗                        ↘ [VAE]
[Positive Conditioning] ↗
[Negative Conditioning] ↗
[Empty Latent] ↗

Advanced Usage

Dual Model Workflow

Use different models for different phases of generation:

1. Connect primary model for initial cycles
2. Connect secondary model (optional)
3. Set "Secondary Model Start Cycle" to switch models mid-process

Additive Conditioning

Enhance your prompts with additional conditioning:

1. Connect main positive/negative conditioning
2. Connect additive positive/negative conditioning (optional)
3. Adjust strength values and scaling options

High-Resolution Generation

For very high-resolution outputs:

1. Start with lower resolution latent
2. Set higher upscale factor (2.0-4.0)
3. Use more cycles (4-8) for gradual upscaling
4. Enable tiled VAE for memory efficiency

Steps Scaling Strategies

Decrement Mode (Refinement Strategy)

Best for progressive refinement with fewer steps in later cycles:

• Use Case: When you want detailed initial generation, then refinement
• Configuration:
  • Steps Adjustment Mode: decrement
  • Steps Scaling Value: 3-8 (moderate reduction)
  • Threshold Mode: auto (prevents going too low)
• Example: 20 → 15 → 10 → 5 steps across 4 cycles

Increment Mode (Detail Enhancement Strategy)

Best for progressive detail enhancement with more steps in later cycles:

• Use Case: When you want quick initial generation, then detailed refinement
• Configuration:
  • Steps Adjustment Mode: increment
  • Steps Scaling Value: 5-10 (moderate increase)
  • Threshold Mode: manual with reasonable cap (e.g., 50)
• Example: 10 → 15 → 20 → 25 steps across 4 cycles

Gradual Upscaling

For smoother upscaling transitions:

1. Enable "Gradual Upscaling"
2. Set "Gradual Upscale Steps" to 3-5
3. Use with higher upscale factors (3.0+) for best results

Post-Processing Nodes

ComfyUI-KarmaNodes includes specialized post-processing nodes for enhancing your generated images with professional-grade effects and color corrections.

🎬 Karma Film Grain

Add authentic analog film grain texture to your images for a cinematic, vintage aesthetic.

Features

• Realistic Multi-Layer Grain: Combines multiple noise layers for authentic film texture
• Luminance-Based Intensity: More grain in darker areas, mimicking real film behavior
• Configurable Parameters: Adjustable grain strength and particle size
• Reproducible Results: Seed-based grain patterns for consistent outputs
• Fallback Support: Works even without scipy dependency

Parameters

• Strength (0.01-1.0): Intensity of the film grain effect
• Grain Size (0.1-5.0): Size/scale of grain particles
• Seed (0-2³¹-1): Random seed for reproducible grain patterns

Usage Tips

• Subtle Effects: Use strength values 0.05-0.15 for realistic film look
• Vintage Style: Higher strength (0.2-0.4) with larger grain size (2.0-3.0)
• Fine Detail: Smaller grain size (0.5-1.0) for high-resolution images
• Consistency: Use the same seed across similar images for matching grain patterns

🎨 Karma Kolors

Professional color grading and correction tools for precise image enhancement.

Features

• White Balance Correction: Temperature-based color correction (2000K-10000K)
• Brightness Control: Precise brightness adjustments (-20% to +20%)
• Contrast Enhancement: Professional contrast control with fine increments
• Saturation Adjustment: Color intensity control for vibrant or muted looks
• Optimal Processing Order: Adjustments applied in the correct sequence for best results

Parameters

• White Balance: Temperature in Kelvin (2000K-10000K) or "auto"
  • 2000K-3000K: Warm, candlelight tones
  • 3000K-4000K: Warm white, tungsten lighting
  • 5000K-6500K: Daylight, neutral white
  • 6500K-10000K: Cool, blue-tinted lighting
• Brightness (-20% to +20%): Overall image brightness in 0.5% increments
• Contrast (-20% to +20%): Contrast adjustment in 0.5% increments
• Saturation (-20% to +20%): Color intensity in 0.5% increments

Usage Tips

• Natural Corrections: Start with small adjustments (±2-5%)
• Creative Looks: Combine temperature shifts with saturation changes
• Portrait Enhancement: Slight brightness (+2-5%) and contrast (+3-8%)
• Landscape Vibrancy: Increase saturation (+5-10%) with slight contrast boost
• Vintage Look: Warm temperature (3000K-4000K) with reduced saturation (-5-10%)

Post-Processing Workflow Examples

Basic Enhancement Chain

[Generated Image] → [Karma Kolors] → [Karma Film Grain] → [Save Image]

Professional Color Grading

[Generated Image] → [Karma Kolors] → [Additional Processing] → [Final Output]
                     ↓
                   White Balance: 5500K
                   Brightness: +3.0%
                   Contrast: +5.0%
                   Saturation: +2.0%

Cinematic Film Look

[Generated Image] → [Karma Kolors] → [Karma Film Grain] → [Save Image]
                     ↓                ↓
                   Warm tone (3200K)   Strength: 0.12
                   Contrast: +8%       Grain Size: 1.5
                   Saturation: -3%     Seed: 42

Requirements

• Python: 3.9+
• PyTorch: 2.0.0+
• Pillow: 9.0.0+
• NumPy: 1.21.0+
• scikit-image: 0.19.0+
• ComfyUI: Latest version

Optional Dependencies

• scipy: 1.9.0+ (recommended for enhanced film grain quality)
• colorsys: Built-in Python module (used for color space conversions)

Performance Tips

1. Memory Management: Enable tiled VAE for large images
2. Cycle Count: More cycles = better quality but longer processing time
3. Upscale Factor: Higher factors require more VRAM
4. Model Selection: Use appropriate models for your target resolution

Troubleshooting

Common Issues

Out of Memory Errors:

• Enable tiled VAE processing
• Reduce upscale factor or number of cycles
• Use smaller initial latent size
• Disable gradual upscaling to reduce intermediate steps

Poor Quality Results:

• Increase number of cycles
• Adjust denoise strengths and enable denoise scaling
• Try different upscaling methods
• Enable sharpening filter
• Use steps scaling with increment mode for more detail in later cycles
• Enable gradual upscaling for smoother transitions

Slow Performance:

• Reduce number of cycles
• Use basic upscaling instead of model-based
• Use steps scaling with decrement mode to reduce steps in later cycles
• Disable gradual upscaling
• Use auto threshold mode for steps scaling

Steps Scaling Issues:

• Steps going too low: Use manual threshold mode with appropriate minimum
• Steps going too high: Use manual threshold mode with reasonable maximum
• Inconsistent results: Try auto threshold mode for balanced scaling

Post-Processing Issues:

• Film grain too strong: Reduce strength value (try 0.05-0.10 for subtle effects)
• Grain pattern inconsistent: Use the same seed value across related images
• Color corrections too harsh: Use smaller adjustment increments (±1-3%)
• White balance not working: Ensure temperature value is within 2000K-10000K range
• Scipy warning for film grain: Install scipy for better grain quality: pip install scipy

Contributing

Contributions are welcome! Please feel free to submit issues, feature requests, or pull requests.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Built for the ComfyUI ecosystem
• Inspired by progressive sampling techniques in diffusion models
• Thanks to the ComfyUI community for feedback and testing

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: karma@karmaviz.biz

☕ Support Development

If you find ComfyUI-KarmaNodes useful and want to support future development, consider buying me a coffee! Your support helps maintain and improve these tools, develop new features, and keep everything free and open-source.

☕ Buy me a coffee

Every contribution, no matter how small, is greatly appreciated and directly contributes to:

• 🚀 New node development and features
• 🐛 Bug fixes and performance improvements
• 📚 Better documentation and tutorials
• 🔧 Ongoing maintenance and support

Thank you for being part of the ComfyUI-KarmaNodes community! 🙏

---

Note: This node is designed for advanced users familiar with ComfyUI workflows. Basic knowledge of diffusion model sampling is recommended.