1. Workflow Overview

• Purpose:
  A professional AI portrait photography workflow integrating:

  • Face segmentation (Florence2 + SAM2).

  • Multi-ControlNet (pose + depth).

  • IPAdapter style transfer.

  • Local redraw optimization.

2. Key Nodes

• Florence2Run: Detects faces and outputs masks/BBox.

• Sam2Segmentation: Refines hair/clothing edges.

• ControlNetApplySD3: Dual control (pose 0.6 + depth 1.0).

• IPAdapterAdvanced: Blends style from reference images (strength 0.7).

• GrowMaskWithBlur: Expands mask edges for natural blending.

3. Workflow Structure

1. Face Detection:
   Input image → Florence2Run → BBox/mask.

2. Matting:
   Sam2Segmentation → GrowMaskWithBlur.

3. ControlNet:
   Canny edge + depth control.

4. IPAdapter:
   Style transfer from reference image.

5. Generation:
   KSampler (dpmpp_sde, 20 steps).

6. Post-Processing:
   Redraw (denoise=0.1) + comparison.

4. Inputs & Outputs

• Inputs: Portrait image + style reference.

• Outputs: High-res portrait (PNG) with comparison.

5. Notes

• VRAM: ≥16GB GPU recommended.

• Errors:

  • Check Florence2/SAM2 model paths if masks fail.

• Optimization:

  • Reduce sampler steps or use FP16 models.