BRIA 2.3

🚀 BRIA 2.3: Text-to-Image Model for Commercial Licensing

BRIA AI 2.3 is a revolutionary text-to-image model trained on legal data. It combines technological innovation with ethical responsibility and legal security, setting a new standard in the AI industry. BRIA AI licenses the foundation model with full legal liability coverage, and its dataset is free of copyrighted materials.

🚀 Quick Start

Get Access

BRIA 2.3 is accessible in multiple ways:

• API Endpoint: Bria.ai, fal.ai
• ComfyUI: Use it in workflows

For Commercial Use

• Purchase: Click Here to obtain a commercial license.

✨ Features

• Legally Compliant: Offers full legal liability coverage for copyright and privacy infringements. Trained on 100% licensed data from leading partners, ensuring ethical content use.
• Patented Attribution Engine: Compensates data partners using proprietary and patented algorithms.
• Enterprise-Ready: Designed for business applications, delivering high-quality, compliant imagery for commercial needs.
• Customizable Technology: Provides access to source code and weights for extensive customization.

📦 Installation

To install the necessary library for using the model, run the following command:

pip install diffusers

💻 Usage Examples

Basic Usage

from diffusers import DiffusionPipeline
import torch

pipe = DiffusionPipeline.from_pretrained("briaai/BRIA-2.3", torch_dtype=torch.float16, use_safetensors=True)
pipe.force_zeros_for_empty_prompt = False
pipe.to("cuda")

prompt = "A portrait of a Beautiful and playful ethereal singer, golden designs, highly detailed, blurry background"
negative_prompt = "Logo,Watermark,Text,Ugly,Morbid,Extra fingers,Poorly drawn hands,Mutation,Blurry,Extra limbs,Gross proportions,Missing arms,Mutated hands,Long neck,Duplicate,Mutilated,Mutilated hands,Poorly drawn face,Deformed,Bad anatomy,Cloned face,Malformed limbs,Missing legs,Too many fingers"

images = pipe(prompt=prompt, negative_prompt=negative_prompt, height=1024, width=1024).images[0]

📚 Documentation

Model Description

Tips for Inference

⚠️ Important Note

• You must set pipe.force_zeros_for_empty_prompt = False.
• Using negative prompt is recommended.
• For fine-tuning, use null text embedding instead of zeros (see SD 1.5 fine-tuning scripts for example).
• The model supports multiple aspect ratios, with a total resolution of approximately 1024*1024 = 1M pixels, e.g., (1024,1024), (1280, 768), (1344, 768), (832, 1216), (1152, 832), (1216, 832), (960,1088).
• If you need speed, try the BRIA 2.3 Fast, which achieves comparable quality with 75% reduced inference time.
• For BRIA 2.3 (not Fast), use 30 - 50 steps (higher is better); the Fast model works well with just 8 steps.
• For BRIA 2.3 (not Fast), use guidance_scale of 5.0 or 7.5; for the Fast models, use 1.0.

📄 License

This model is open-source for non-commercial use. Purchase is required to license and access the model for commercial use. Commercial licensing terms & conditions.

Additional Information

• New Version: Bria 3 generation is available: Bria 3.1
• What's New: BRIA AI 2.3 improves the generation of illustration, art, and human faces compared to BRIA 2.2. CLICK HERE FOR A DEMO
• Other Models: Check out our entire family of models, including our Fast model and our HD model
• Compatible Auxiliary Models:
  • ControlNet BG Generation
  • ControlNet Gen Fill
  • ControlNet Inpainting
  • ID Preservation Adapter
  • LoRA LCM
  • IP Adapter
  • ControlNet Pose
  • ControlNet Recoloring
  • ControlNet CannyEdges
  • ControlNet Depth
  • ControlNet Pose
• Community: Join our Discord community for more information, tutorials, tools, and to connect with other users!