text2image prompt generator

Introduction

The Text2Image-Prompt-Generator is a GPT-2 model fine-tuned on the succinctly/midjourney-prompts dataset. This dataset includes 250,000 text prompts collected from users of the Midjourney text-to-image service over a one-month period. The model is designed to auto-complete prompts for any text-to-image model, including those in the DALL·E family.

Architecture

This prompt generator is based on the GPT-2 architecture, fine-tuned to handle text prompts specific to the text-to-image transformation domain. It may produce Midjourney-specific tags and can be adjusted using parameters like aspect ratio or entity importance.

Training

The model is trained on the succinctly/midjourney-prompts dataset, which is a comprehensive collection of text prompts. This dataset captures the variety and complexity of user inputs to the Midjourney service, allowing the model to generate contextually relevant and creative text prompts for image generation.

Guide: Running Locally

To run the Text2Image-Prompt-Generator locally, follow these steps:

1. Clone the Repository: Download the model files from the Hugging Face repository.
2. Install Requirements: Ensure you have Python and PyTorch installed. Use pip install transformers to install the necessary libraries.
3. Load the Model: Use the transformers library to load the model and tokenizer.
4. Generate Prompts: Input your initial text prompt and let the model generate completions.

For optimal performance, consider using a cloud GPU service such as AWS EC2, Google Cloud, or Azure to handle the computational requirements efficiently.

License

The Text2Image-Prompt-Generator model is licensed under the Creative Commons Attribution 2.0 (cc-by-2.0). When using this model, credit must be attributed to Succinctly AI.