1. Workflow Overview

This workflow is designed to automatically beautify portrait photos through AI models, featuring functions like skin smoothing, acne removal, hair refinement, and background blurring. It primarily uses the majicMIX realistic model for realistic-style enhancements, with Ultralytics detectors and SAM models for facial region localization and detailing.

2. Core Models

• majicMIX realistic v2: Main model for photorealistic portraits (download to ComfyUI/models/checkpoints).

• YOLOv8m-seg: Ultralytics segmentation model for face detection (requires comfyui-detection plugin).

• SAM (Segment Anything Model): Meta's model for mask refinement (download sam_vit_b_01ec64.pth).

3. Key Nodes

4. Workflow Structure

1. Model Loading Group:

   • Input: None

   • Output: Model, CLIP, VAE

   • Nodes: CheckpointLoaderSimple

2. Face Detection & Masking:

   • Input: Raw image

   • Output: Face BBox and masks

   • Nodes: UltralyticsDetectorProvider + SAMLoader

3. Beautification Group:

   • Input: Image + prompts (e.g., "smooth skin, 8K")

   • Output: Enhanced image

   • Nodes: FaceDetailer + CLIPTextEncode

5. Inputs & Outputs

• Inputs:

  • Image (e.g., 87979.jpg)

  • Positive prompt: "Smooth skin, refine, remove acne, HD"

  • Negative prompt: "Split hair, large pores"

• Outputs: Saved image in ComfyUI folder.

6. Notes

• Requires ≥6GB VRAM (GPU recommended).

• Effects may be subtle; adjust prompts/parameters (e.g., denoise=0.5) manually.

• Dependencies: ComfyUI-Impact-Pack, comfyui-detection.