1. Workflow Overview

This workflow converts real photos into minimalist gradient-style illustrations by:

• Triple ControlNet (Canny/Depth/Scribble) for structure preservation

• IPAdapter style transfer for gradient color effects

• Chibi LoRA enhancing cartoonish style

• Outputs HD illustrations at 896x1184 resolution

2. Core Models

3. Key Nodes

4. Workflow Structure

• Group 1: Input Processing

  • LoadImage: Resizes input to 896x1184

  • Canny/Depth/XDoG: Generates control maps

• Group 2: Style Transfer

  • IPAdapter: Injects gradient style (prompt: "Flat gradient illustration, beige background")

  • LoRA Loader: Applies Chibi style

• Group 3: Generation

  • ControlNetApplyAdvanced: Multi-control (edge=0.4/depth=0.2/scribble=0.2)

  • KSampler: Uses DPM++ 2M (30 steps, CFG=7)

5. Inputs & Outputs

• Required Inputs:

  • Any real photo (clear subject recommended)

  • Style prompt (e.g., "Clean modern design with flat gradients")

• Outputs:

  • PNG illustration (896x1184)

  • Auto-saved to ComfyUI/output

6. Notes

⚠️ Critical Configs:

1. VRAM Optimization:

   • Enable VAEDecodeTiled for 8GB GPUs

   • Reduce resolution to 768x1024 if needed

2. Plugin Dependencies:

   # ControlNet
   git clone https://github.com/comfyanonymous/ComfyUI-ControlNet
   # IPAdapter
   git clone https://github.com/cubiq/ComfyUI_IPAdapter_plus
   # Scribble
   git clone https://github.com/space-nuko/ComfyUI-ScribbleNodes

3. Style Tuning:

   • Adjust weight in IPAdapterAdvanced (0.3-0.6)

   • Modify threshold in XDoG_Preprocessor (default=30)