coloringbookflux

Introduction

Coloring Book Flux is a text-to-image model designed for generating simple, cute illustrations suitable for coloring books and other creative uses. It leverages LoRA (Low-Rank Adaptation) technology and is trained with the Flux framework, optimized for creating black-and-white line art ideal for coloring.

Architecture

The model is based on a Flux LoRA architecture, using a synthetic dataset of 100 images featuring human, vehicle, and animal illustrations. The images are captioned using Joy Caption Batch. Adjustments like limiting repeats, increasing epochs, and lowering DIM/ALPHA settings were applied to enhance performance.

Training

The training process involved numerous iterations to refine the model, focusing on achieving consistent results. The DEIS sampler is recommended for optimal performance. The model supports generating both black-and-white and colored images. The training included strategies to minimize subject blending and improve style consistency.

Guide: Running Locally

To run the Coloring Book Flux model locally, follow these steps:

1. Install Dependencies: Ensure you have Python and PyTorch installed. Use pip install diffusers to install the necessary library.
2. Clone the Repository: Clone the model repository from Hugging Face.
3. Download Model Weights: Access the Files & versions tab to download model weights in Safetensors format.
4. Load the Model: Use the Hugging Face Transformers library to load and initialize the model with the downloaded weights.
5. Generate Images: Use prompts like "c0l0ringb00k" to generate images. Adjust prompts for different styles and subjects.

For better performance, consider using cloud GPUs such as AWS EC2 with NVIDIA GPUs or Google Cloud's AI Platform.

License

The Coloring Book Flux model is licensed under CreativeML Open RAIL-M. This license allows for open use with certain restrictions, especially concerning harmful applications. Be sure to review the full license terms before use.