/AI-Selfie-with-Celebrities-Generator

Primary LanguagePython

AI Selfie with Celebrities Generator

An AI-powered tool that generates realistic selfies between fans and celebrities using Google's Gemini AI model.

Features

  • Generate authentic-looking selfies with celebrities
  • Uses advanced AI image generation with style transfer
  • Automatic verification system to ensure quality
  • Supports JPEG input images
  • High-quality output with realistic lighting and perspective

Requirements

  • Python 3.7+
  • Google Gemini API key
  • Required Python packages:
    • google-genai
    • Pillow (PIL)

Installation

  1. Clone this repository:
git clone <repository-url>
cd AI-Selfie-with-Celebrities-Generator
  1. Install required packages:
pip install google-genai Pillow
  1. Set up your Gemini API key:
export GEMINI_API_KEY="your-api-key-here"

Usage

  1. Place your celebrity photo as kamal.jpeg
  2. Place your fan photo as fan.jpeg
  3. Run the script:
python main.py

The generated celebrity selfie will be saved as celebrity_selfie.png.

How It Works

  1. Image Loading: Loads the celebrity photo (kamal.jpeg) and fan photo (fan.jpeg)
  2. AI Generation: Uses Gemini 2.5 Flash model to create a realistic selfie
  3. Quality Check: Compares input and output to ensure transformation occurred
  4. Verification: AI analyzes the generated image for authenticity and realism
  5. Output: Saves the final celebrity selfie

Output Quality

The system includes automatic verification that rates:

  • Visibility and recognition of both individuals (1-10)
  • Natural perspective and realism (1-10)
  • Consistent lighting and shadows (1-10)
  • Appropriate expressions (1-10)

Typical output scores 9.5/10 for overall authenticity.

File Structure

AI-Selfie-with-Celebrities-Generator/
├── main.py                 # Main application script
├── kamal.jpeg             # Celebrity photo
├── fan.jpeg               # Fan photo
├── celebrity_selfie.png   # Generated output
└── README.md              # This file

API Configuration

The script uses Google's Gemini 2.5 Flash Image Preview model with:

  • Seed: 42 (for consistent results)
  • MIME type: image/jpeg for inputs
  • Output format: PNG

Troubleshooting

  • Ensure your GEMINI_API_KEY environment variable is set correctly
  • Check that input images are in JPEG format
  • Verify both kamal.jpeg and fan.jpeg exist in the project directory

License

This project is open source and available under the MIT License.