Flux Uncensored V2

Introduction

The FLUX Uncensored LoRA V2 is an updated model by EnhanceAI designed for generating high-quality, uncensored images using the base model black-forest-labs/FLUX.1-dev. This version focuses on improved performance, stability, and reduced model size, optimized for generating explicit content from user prompts. It is intended for research and personal use under a non-commercial license.

Architecture

The model utilizes the LoRA (Low-Rank Adaptation) technique to enhance the capabilities of the base model by fine-tuning the weights to eliminate content restrictions. It is integrated into the FLUX.1-dev pipeline using the diffusers library, enabling the generation of NSFW (Not Safe For Work) content.

Training

The FLUX Uncensored LoRA V2 is built on black-forest-labs/FLUX.1-dev, with updates for performance and stability. The LoRA weights are streamlined to maintain quality while reducing the model's size, ensuring faster and more reliable outputs.

Guide: Running Locally

1. Prerequisites:

   • Ensure you have a CUDA-enabled GPU for optimal performance.
   • Install the diffusers library and other dependencies.

2. Basic Steps:

   from diffusers import AutoPipelineForText2Image
   import torch
   
   # Load the base model
   pipeline = AutoPipelineForText2Image.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16).to('cuda')
   
   # Load the uncensored LoRA weights
   pipeline.load_lora_weights('enhanceaiteam/Flux-uncensored-v2', weight_name='lora.safetensors')
   
   # Generate an image with an uncensored NSFW prompt
   image = pipeline('a naked cute girl').images[0]
   image.show()
   

3. Cloud GPUs: Consider using cloud services like AWS, Google Cloud, or Azure for access to powerful GPUs.

License

The model is distributed under the FLUX-1-dev Non-Commercial License. You can find more details and the full license text here. The license allows for research and personal use only, prohibiting commercial exploitation.