AI Cooking Coach

AI Cooking Coach is a Python application that uses Google's Gemini Pro Vision to analyze your cooking videos and provide step-by-step feedback, just like a personal cooking instructor. It generates a new video with feedback overlaid directly on your original footage.

---

🚀 Features

• AI-Powered Analysis: Leverages the Gemini API's multimodal capabilities to understand and comment on cooking techniques shown in a video.
• Customizable Persona: Choose your coach's style—from Gordon Ramsay's tough love to Julia Child's encouragement—by editing the prompt.
• Video Feedback Overlay: Feedback appears at the precise moments actions occur in your video.
• Audio Preservation: The original audio is included in the final output.
• Secure & Structured: API keys are managed securely using environment variables.

---

🧑‍🍳 How It Works

1. AI Analysis (analyzer.py)

• The script sends your video and a detailed prompt to the Gemini API.
• The AI analyzes the video and returns a structured JSON file with timestamped feedback.

Sample Output:

{
  "analysis": [
    {
      "timestamp": "00:14-00:15",
      "action": "Cracking eggs",
      "feedback": "Technique is acceptable, but lacks precision. Crack the eggs on a flat surface to avoid shell fragments."
    },
    {
      "timestamp": "00:16-00:18",
      "action": "Seasoning",
      "feedback": "Good amount of salt and pepper, but seasoning before whisking can make the eggs watery. It's better to season in the pan."
    }
  ],
  "summary": {
    "title": "Analysis of My Cooking",
    "overall_feedback": "A solid attempt. Your fundamentals are there, but focus on refining your technique for a more professional result. Pay attention to heat management and the timing of your seasoning."
  }
}

2. Video Visualization (visualizer.py)

• Reads the generated JSON and original video.
• Uses OpenCV to draw feedback text onto each frame at the correct timestamp.
• Uses moviepy to combine the new video track with the original audio.

---

💡 Feedback Examples

• "Technique is acceptable, but lacks precision. Crack the eggs on a flat surface to avoid shell fragments."
• "Good amount of salt and pepper, but seasoning before whisking can make the eggs watery. It's better to season in the pan."

---

🛠️ Setup and Installation

1. Clone the Repository

git clone <your-repo-url>
cd ai-cooking-coach

2. Create a Virtual Environment

For Mac/Linux:

python3 -m venv venv
source venv/bin/activate

For Windows:

python -m venv venv
.\venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Set Up Your API Key

• Create a file named .env in the root directory.
• Add your API key:

  GEMINI_API_KEY="YOUR_API_KEY_HERE"

• The .gitignore is already configured to keep this file private.

---

▶️ How to Run

1. Place your video: Put the cooking video you want to analyze into the videos/input/ directory.
2. Run the main script:

   python main.py "videos/input/your_video_name.mp4"

3. Check the output:
   • The AI's analysis will be saved as a JSON file in the analysis_results/ folder.
   • The final video with feedback overlay will be saved in the videos/output/ folder.

---

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

Build your own AI-powered cooking coach or integrate it into your product!