1. Workflow Overview

This workflow generates 3D Q-style children's illustrations for storybooks or animations, featuring:

• Stylized Outputs: Enhanced by LoRAs (e.g., FLUX-Cute 3D Kids Style).

• Upscaling: UltimateSDUpscale for 4K resolution.

• Multi-LoRA Fusion: Combines base model with auxiliary LoRAs for details.

• Default Theme: "Space-traveling polar bear" (customizable prompts).

2. Core Models

3. Key Components

4. Workflow Structure

Group 1: Model Loading

• Inputs: None (auto-loads models).

• Outputs: Initialized models + LoRAs.

• Key Nodes: UNETLoader, LoraLoader.

Group 2: Text-to-Image Generation

• Inputs: English prompts (e.g., "A polar bear in space").

• Outputs: Initial image (768x1024).

• Key Nodes: CLIPTextEncode, KSampler (Euler).

Group 3: Upscaling (Optional)

• Inputs: Low-res generated image.

• Outputs: 4K image.

• Key Nodes: UltimateSDUpscale, 4x-AnimeSharp.

5. Inputs & Outputs

• Inputs:

  • Prompt (default: English, editable).

  • Seed (random/fixed).

  • Resolution (default: 768x1024).

• Outputs:

  • Initial image (saved via SaveImage).

  • Upscaled image (enable UltimateSDUpscale).

6. Notes

1. LoRA Models:

   • Download from links in Note nodes and place in models/loras.

2. VRAM:

   • 8GB+ for generation, 12GB+ for upscaling.

3. Optimization:

   • Adjust tile_size (default: 64) in UltimateSDUpscale for lower VRAM.

4. Dependencies:

   # Install plugins
   cd ComfyUI/custom_nodes
   git clone https://github.com/pysssss/comfyui-custom-scripts.git
   git clone https://github.com/ssitu/ComfyUI-UltimateSDUpscale.git