GPT Image 1 – Prompt-based Image Transformation

This project demonstrates the use of OpenAI's GPT Image 1 model to generate new images from existing ones using two distinct prompts.

📌 Overview

• ✅ Input: 10 original images
• 🔁 Process: Each image processed with 2 different prompts
• 🖼️ Output: 20 AI-generated images (2 per input)
• 🕒 Logged: Inference time per image
• 📊 Analyzed:
  • Average processing time
  • Estimated throughput (images per minute/hour)
  • Cost per image and total project cost
  • Impact of input image size on speed

📂 Files

• /outputs/ — folder with all generated images (*_v1.png, *_v2.png)
• timing_log.csv — duration of each inference
• report_summary.txt — summary of findings (speed, cost, scaling limits)

💡 Key Findings

• Average image generation time: ~30.2 seconds
• Estimated throughput: ~2 images/min or 119 images/hour
• Cost per image (medium quality, 1536×1024): $0.063
• Total cost for 20 images: $1.26
• Input image size affects processing speed
• API Tier level (IPM limit) may throttle real-world performance

🛠️ Tech Stack

• OpenAI GPT Image 1 API
• Python (PIL, pandas)
• Streamlit (planned for frontend demo)

📎 Report & Results

• 📄 report_summary.md
• 📁 [Output folder with images & logs]