Wizards_vintage_romance_novel F L U X

Introduction

The Wizards Vintage Romance Novel-FLUX is a bespoke text-to-image model designed for generating vintage romance book cover illustrations. It uses the LoRA (Low-Rank Adaptation) technique and is optimized for the FLUX environment.

Architecture

The model is based on the black-forest-labs/FLUX.1-dev architecture and utilizes LoRA for enhancing image generation quality. It requires specific prompts to generate images, particularly focusing on Harlequin romance book covers with pulp art styles.

Training

The model was trained using specific settings, including Hyper 8-step LoRA and 10-12 steps DDIM Trailing (SGM Uniform) with a CFG scale of 3.5. It is case-sensitive to some extent, meaning prompts should be crafted carefully to ensure high-quality outputs.

Guide: Running Locally

1. Setup Environment: Ensure you have Python installed along with PyTorch and the Hugging Face Diffusers library.
2. GPU Requirement: For optimal performance, use a CUDA-enabled GPU. Cloud GPUs such as AWS, Google Cloud, or Azure are recommended.
3. Install Dependencies:

   pip install torch diffusers
   

4. Load the Model:

   from diffusers import AutoPipelineForText2Image
   import torch
   
   device = "cuda" if torch.cuda.is_available() else "cpu"
   pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.bfloat16).to(device)
   pipeline.load_lora_weights('WizWhite/wizard-s-vintage-romance-novel', weight_name='Wiz-Vintage_Romance-FLUX.safetensors')
   

5. Generate Image:

   image = pipeline('Photo of a harlequin romance book cover with the title "The Wizard's Vintage Romance". The book is on a bedside table. It features a pulp art illustration of Wizard Whitebeard with a long white beard. Promotional sticker on the bottom say "THE ORIGINAL - now for FLUX"').images[0]
   

License

The model is distributed under a bespoke LoRA-trained license, allowing commercial use and derivatives, with restrictions on using different licenses. More details can be found here.