Add Stable Diffusion portrait generation script

[banti0238-afk]

Overview

Adds a standalone Python script to generate portrait images using Stable Diffusion v1.5 with CPU-optimized inference settings.

Changes

• generate_portrait.py: New script for generating portraits with the following features:
  • Uses the Hugging Face diffusers library for easy model loading and inference
  • Optimized for CPU execution with attention slicing and reduced inference steps (20 instead of default 50)
  • Disables safety checker to improve performance on CPU-constrained environments
  • Generates 512x512 images using a professional photography prompt template
  • Automatically creates output directory and saves results as PNG

Implementation Details

• Model: runwayml/stable-diffusion-v1-5 from Hugging Face Hub
• Inference steps: 20 (reduced from default 50 for faster CPU generation)
• Guidance scale: 7.5
• Image size: 512x512
• Output: outputs/hot_woman_portrait.png

Usage

python generate_portrait.py

The script will automatically download the model on first run and generate the image. Typical CPU generation time: 15-30 minutes.

Notes

• Requires: torch, diffusers, transformers, pillow, scipy, safetensors
• GPU (CUDA) detection supported for faster inference when available
• Memory optimized with attention slicing for systems with limited resources