mooniverse

Introduction

The MOONIVERSE model by Alvdansen is a text-to-image generator that uses the Stable Diffusion framework and incorporates LoRA (Low-Rank Adaptation) techniques. It is designed to produce aesthetically pleasing, surreal images with natural textures and muted tones.

Architecture

MOONIVERSE is based on the black-forest-labs/FLUX.1-dev base model. It employs the Diffusers library, which is optimized for generating high-quality images from textual prompts. The model exhibits a bias towards East Asian portraits but can be guided towards other features using specific prompt descriptors and adjusting the LoRA weight.

Training

The model training was sponsored by Replicate. It was fine-tuned with a focus on creating images with a surreal style, leveraging the capabilities of LoRA for enhanced adaptability in generating diverse image styles.

Guide: Running Locally

1. Clone the Repository: Begin by cloning the MOONIVERSE repository from Hugging Face.
2. Set Up Environment: Ensure your environment includes Python and the necessary libraries, such as Hugging Face's Transformers and Diffusers.
3. Download Weights: Access the Files & Versions tab to download the Safetensors format weights.
4. Run the Model: Use the downloaded weights with a text-to-image pipeline to generate images. Include the "surreal style" trigger word in your prompts for optimal results.

Suggested Cloud GPUs

For high-performance requirements, consider using cloud services that offer GPUs, such as AWS EC2 with NVIDIA GPUs, Google Cloud Platform's AI Platform, or Azure's GPU-powered virtual machines.

License

MOONIVERSE is released under the Creative Commons Attribution-NonCommercial 3.0 (cc-by-nc-3.0) license. This license allows for modification and sharing under similar terms, but it is not intended for commercial use unless explicitly granted.