1. Workflow Overview

This workflow specializes in converting daytime scenes to nighttime, ideal for architectural visualization and post-production. Key features:

• Lighting Transformation: Preserves composition via ControlNet while adding night effects (e.g., blue sky, warm indoor lights).

• Style Control: Enhances night elements with prompts like (Night view:1.2).

• High-Resolution Output: Supports up to 2048px with optional upscaling.

Use Cases: Architectural renders, film VFX, game environment design.

2. Core Models

3. Key Nodes

4. Workflow Structure

1. Input Preprocessing

   • Nodes: LoadImage → ImageResize+

   • Input: Daytime image (auto-resized to 2048px width).

   • Output: Normalized image.

2. Conditioning Encoding

   • Nodes: CLIPTextEncode + easy stylesSelector

   • Input: Night prompts + Fooocus negative presets.

   • Output: Text embeddings.

3. ControlNet Processing

   • Nodes: ACN_ControlNet++LoaderSingle → ACN_AdvancedControlNetApply

   • Params: Weight=0.9, start=0%, end=80%.

   • Output: Conditioned signals with composition constraints.

4. Image Generation

   • Nodes: KSampler Adv. ×2 (two-phase generation).

   • Params: CFG=3.5, seed=369650626139512, 1-step draft + 20-step refine.

   • Output: Night image (comparison via Image Comparer).

5. Post-Processing

   • Node: SaveImage

   • Output Path: ComfyUI/output/ with metadata.

5. Inputs & Outputs

• Required Inputs:

  • Image: Daytime scene (recommended ≥1920x1080).

  • Prompts: Must include night keywords (e.g., "blue sky").

• Output:

  • Image: PNG with comparison (original vs. night).

6. Notes

⚠️ Common Issues:

• VRAM Overflow: Reduce resolution to 1024px or enable torch.compile.

• ControlNet Failure: Verify model matches (controlnet-union-sdxl).

• Unrealistic Lighting: Tweak prompt weights (e.g., (Night view:1.5)) or CFG (3-5).

🔧 Optimization Tips:

• Use Seed Everywhere for reproducible results.

• Lower draft steps to 1 for faster testing.