1. Workflow Overview

This e-commerce product enhancement workflow integrates three core functions:

1. Auto Background Replacement: Isolate products from white-background images (e.g., cosmetics → forest/lab scenes)

2. IC Lighting Simulation: Physically accurate lighting via ICLight model

3. 4K Upscaling: Detail-preserving super-resolution (4x_NMKD-Siax_200k)

2. Core Models

3. Key Nodes

4. Workflow Groups

1. Input Group

   • Nodes: LoadImage (product + style reference)

   • Input: 1080x1080 white-background product image

2. Matting & Preprocessing

   • Nodes: SegmentAnythingUltra → GrowMaskWithBlur

   • Output: Feather-edged product mask

3. Background Generation

   • Nodes: KSampler + BrushNet

   • Params:

     • Prompt: "masterpiece, RAW photo, 8k details"

     • LoRA: E-commerce Forest Scene XL (weight 0.5)

4. IC Lighting

   • Nodes: LightSource + ICLightConditioning

   • Params: Top Right Light, strength 0.185

5. Final Output

   • Nodes: UltimateSDUpscale → SaveImage

   • Params: Tile size 512x512, CRF 19

5. Inputs & Outputs

• Inputs:

  • Product image: White/plain background (≥1080x1080)

  • Style prompt: e.g., "lab scene, metallic texture, cool tone"

  • Lighting: Direction/color/intensity (default: top-right white)

• Output:

  • 4K PNG (optional transparency)

  • Metadata includes all generation parameters

6. Notes

1. VRAM Optimization:

   • 6GB VRAM min (8GB+ recommended for upscaling)

   • Enable tile_size=512 to avoid OOM

2. Matting Tips:

   • Adjust GrowMaskWithBlur radius for complex edges (default 1px)

3. Troubleshooting:

   • Unnatural lighting: Reduce ICLightConditioning weight (e.g., 0.1→0.05)

   • Background leaks: Increase BrushNet mask_grow (default 10px)