1. Workflow Overview

This workflow generates Chinese paper-cut style art images, blending AI with traditional aesthetics. Key steps:

1. Uses Flux model framework to create high-res (1024x1536) latent images.

2. Applies LoRA (Paper-cut_v1.0) to enhance paper-cut textures.

3. Outputs hand-painted style artwork with intricate cut-out details.
   Use Case: Cultural design, festival posters, heritage promotion.

2. Core Models

3. Key Nodes

Dependencies:

• Plugin ComfyUI-Flux (GitHub)

• Models Base Algorithm_F.1.safetensors & ae.sft (VAE)

4. Workflow Groups

• Group 1: Model Loading

  • Nodes: UNETLoader → LoraLoaderModelOnly

  • Loads base model + LoRA.

• Group 2: Text Encoding

  • Nodes: DualCLIPLoader → CLIPTextEncode

  • Prompt: "Exquisite hand-painted illustration, beautiful girl, side smile, black hair"

  • FluxGuidance strength: 3.5.

• Group 3: Image Generation

  • Nodes: SamplerCustomAdvanced → SaveImage

  • Sampling: 20 steps, DPM++ 2M, simple scheduler.

  • Output: PNG paper-cut art (saved to ComfyUI/output).

5. Inputs & Outputs

• Inputs:

  • Resolution: Fixed 1024x1536 (portrait).

  • Prompt: Include "paper-cut", "hand-painted".

  • Seed: Random (can be fixed).

• Output:

  • HD paper-cut image (filename prefix ComfyUI).

6. Notes

⚠️ VRAM: Minimum 10GB (high-res generation).
⚠️ Errors:

• Blurry textures → Verify LoRA filename.

• Color shifts → Try alternate VAE models.
  ⚠️ Tips:

• Add "red background", "symmetrical" to prompts.

• Adjust FluxGuidance (3.5-5.0).

Demo Steps

1. Edit prompts in CLIPTextEncode (e.g., "New Year paper-cut, red background").

2. Reduce resolution in EmptySD3LatentImage if needed (e.g., 768x1024).

3. Run Queue Prompt. Output saves automatically.