1. Workflow Overvie

This workflow generates cityscape images with a grand perspective, suitable for cyberpunk or futuristic scenes. It includes two modules: Online Run (default) and Local Run (manual activation).

Core Models:

• Main Model: miluo-flux-pro_模型配置更智能_v1.0 (custom model, download from LibLib.art)

• LoRA: miluo丨辉光城市-宏大视角_D1.03 (style enhancement, weight=0.8)

• VAE: ae.sft (for image decoding)

2. Node Breakdown

Dependencies:

• Model Downloads:

  • Main Model: LibLib.art Link

  • LoRA: LibLib.art Link

• Plugins:

  • ttN nodes: For dynamic text input (manual install).

  • Baidu Translate Plugin: Requires API key (optional).

3. Workflow Groups

1. Online Run Group (Default):

   • Input: Pre-set prompts (e.g., "cyberpunk city at dusk").

   • Output: Quick image generation for testing.

2. Local Run Group (Manual Activation):

   • Input: Custom prompts via ttN text.

   • Process:

     • Text translation → CLIP encoding → LoRA application → Sampling → VAE decode.

   • Output: HD image (768x1248 or 768x1024).

4. Inputs & Outputs

• Input Parameters:

  • Resolution: 768x1248 (Online) or 768x1024 (Local).

  • Sampler: Euler, 30 steps, CFG=1.

  • Prompts:

    • Online: Pre-defined cyberpunk descriptions.

    • Local: Custom EN/CN input (auto-translated).

• Output: PNG images saved to ComfyUI/output.

5. Notes

1. Model Compatibility:

   • Ensure correct paths for main model and LoRA.

2. VRAM: ≥8GB GPU recommended. Reduce resolution if OOM occurs.

3. Errors:

   • Missing Node: Install plugins (e.g., ttN nodes).

4. Optimization: Enable xformers for faster generation.