1. Workflow Overview

This workflow leverages multiple ControlNets for precision image generation, optimized for interior design visualization. Key features:

1. Dual-ControlNet: Depth + Canny for spatial and edge control.

2. Flux Framework: Uses Interior Design Flux FP8 Model (1024x1536 output).

3. LoRA Enhancement: Metallic Typography LoRA for material textures.
   Use Case: Design proposals, furniture layout previews.

2. Core Models

3. Key Nodes

Dependencies:

• Plugins ComfyUI-Flux & comfyui_controlnet_aux

4. Workflow Groups

• Group 1: Model Loading

  • Nodes: CheckpointLoaderSimple → LoRA Stack

  • Loads base model, LoRA, and ControlNets.

• Group 2: Preprocessing

  • Nodes: LoadImage → Dual AIO_Preprocessor (Depth + Canny).

  • Resolution: 1024x1536.

• Group 3: Generation

  • Nodes: ControlNetApplyAdvanced → KSampler (20 steps, Euler).

  • Output: HD interior render (auto-saved).

5. Inputs & Outputs

• Inputs:

  • Reference image (e.g., floor plan, 580x750 recommended).

  • Prompt: "Gilt on a black background, words" (customizable).

• Output:

  • 1024x1536 render (filename prefix ComfyUI).

6. Notes

⚠️ VRAM: Minimum 12GB (dual ControlNets + high-res).
⚠️ Troubleshooting:

• Control failure → Verify ControlNet/preprocessor match.

• Distortion → Adjust strength in ControlNetApplyAdvanced (default:0.8).
  ⚠️ Tips:

• Use DepthAnythingV2Preprocessor for better depth estimation.

• Try dpmpp_2m sampler for finer details.

Demo Steps

1. Replace image in LoadImage (clear contours work best).

2. Edit prompt in CLIPTextEncode (e.g., "Modern living room, warm tones").

3. Run Queue Prompt. Output saves to ComfyUI/output.