🧠🎙️ AI-Powered Recipe Generator from Audio

This project is a Flask-based web application that transforms spoken content—such as cooking videos or food-related audio—into structured, AI-generated recipes using the power of Ollama's local language models.

---

🚀 Features

• 🎧 Audio Download & Transcription
  Automatically downloads audio from a given URL and converts it into text using a pre-warmed transcription model.

• 🧠 AI Recipe Generation
  Uses a large language model (e.g., deepseek-r1:14b) via Ollama to analyze the transcribed text and generate a detailed recipe.

• 🧹 Text Cleanup & Formatting
  Removes unnecessary tags and capitalizes bolded text for improved readability.

• ⚙️ GPU Memory Management
  Efficiently clears GPU memory after model usage to optimize performance.

• 🌐 Web Interface
  A simple Flask server provides endpoints for:

  • Viewing the homepage (/)
  • Submitting transcribed text to generate a recipe (/to-recipe-chat)

---

🧩 Modules & Integrations

• download_audio: Handles downloading audio from a URL.
• convert_audio_to_text: Transcribes audio using a local model.
• ollama_intergration: Interfaces with the Ollama server to generate recipes.
• flask: Powers the web server and API endpoints.
• re: Cleans and formats the generated text.
• requests: Enables client-side interaction with the Flask API.

---

🛠️ How It Works

1. User submits a YouTube URL or audio content.
2. Audio is downloaded and transcribed.
3. Transcribed text is sent to Ollama for recipe generation.
4. Generated recipe is cleaned, formatted, and returned to the user.

---

📦 Installation

1. Clone the repository:

   git clone https://github.com/your-username/recipe-generator.git
   cd recipe-generator
   
   

Install dependencies

pip install -r requirements.txt

Make sure Ollama is running locally

ollama serve

Run the Flask app

webserver.py

🧠 Model Notes

• Transcription uses a lightweight "turbo" model
• Recipe generation uses "deepseek-r1:14b" via Ollama

---

🧼 GPU Cleanup

To prevent memory leaks and optimize performance, the app includes GPU memory cleanup after model usage using Python's gc and PyTorch's cuda.empty_cache().

---

🧪 Example: Client-Side Script

Use this Python script to interact with the Flask API and print the formatted recipe:

import requests
import re

# Prompt user for YouTube video URL
data = input("URL for YouTube video you want to turn into a recipe: ")

# Send request to Flask API
url = "http://localhost:5000/to-recipe-chat"
payload = {"text": data}
response = requests.post(url, json=payload)

# Format bolded text to uppercase
def capitalize_bolded_text(text):
    text = re.sub(r"\*\*(.*?)\*\*", lambda match: match.group(1).upper(), text)
    return text

# Parse and display the recipe
result = response.json()
recipe_text = result["recipe"]
formatted_recipe = capitalize_bolded_text(recipe_text)
print(formatted_recipe)

---

📄 License

MIT License. See LICENSE file for details.

---

🤝 Contributing Isaac Hunacek

Pull requests are welcome! For major changes, please open an issue first to discuss what you'd like to change.

---

Happy cooking with AI! 🍳