fillll / convergence-test

Convergence test tool

This tool uses OpenAI's API to generate and describe images in an iterative process. It generates an image based on a text prompt, describes it, and then uses the description as a new prompt for the next image. This process continues for a specified number of iterations.

Features

• Image Generation: Utilizes OpenAI's DALL-E model to create images from text prompts.
• Image Description: Describes generated images using GPT-4.
• Iterative Process: Generates a series of images, each based on the description of the previous one.
• Continuation Capability: Can continue from the last image in a previously generated series.

Prerequisites

• Python 3.x
• OpenAI API Key

Installation

1. Clone the repository to your local machine.
2. Install the required Python packages:

   pip install openai requests

Usage

1. Set up your OpenAI API key:

   • Save your API key in a text file, e.g., openai.key.

2. Run the script:

   python script_name.py -a [API Key File] -g [Generation Prompt] -f [Output Folder] -d [Description Prompt] -n [Number of Iterations]

   Replace script_name.py with the actual script filename.

   Arguments:

   • -a or --api: The file containing the OpenAI API key.
   • -g or --generate-prompt: Initial prompt for generating images.
   • -f or --folder: Folder where to store the images.
   • -d or --describe-prompt: Prompt for describing images. (Default: "Describe the image in all details.")
   • -n or --number: Number of iterations. (Default: 10)

   Example:

   python script_name.py -a openai.key -g "A futuristic cityscape" -f generated_images -d "Describe the image in all details." -n 5

Important Notes

• The script will create a new folder if it doesn't exist.
• If the folder already contains images, the script will continue from the last numbered image.

Disclaimer

This tool uses the OpenAI API, and usage may incur costs based on your OpenAI usage plan.