BRIA 2.3 FAST

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

BRIA 2.3 FAST is the LCM version of BRIA 2.3. It offers the best combination of quality and latency for the 2.X family. This model is trained on legal data, integrating technological innovation with ethical responsibility and legal security. To obtain legal liability, install the Bria Agent.

🚀 Quick Start

Get Access

Bria 2.3Fast is accessible in various ways, including source - code and weights, ComfyUI nodes, or API endpoints.

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

For Commercial Use

• Purchase: For a commercial license, simply click Here.

For more information, visit our website. Join our Discord community for tutorials, tools, and to connect with other users!

✨ Features

What's New

BRIA 2.3 FAST is a high - speed version of BRIA 2.3, offering an optimal balance between speed and accuracy. It takes only 1.64 seconds to generate images on a standard NVIDIA A10 GPU, achieving excellent image quality with an 80% reduction in inference time.

The model is distilled using the LCM technique and supports multiple aspect ratios, with a default resolution of 1024x1024. Similar to Bria AI 2.3, it has improved realism and aesthetics.

Our evaluations show that it achieves comparable image quality to its teacher, BRIA 2.3, and outperforms the SDXL LCM. While SDXL Turbo is faster, our model produces significantly better human faces as it supports higher resolution. These assessments are based on human preferences.

CLICK HERE FOR A DEMO

Key Features

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

📚 Documentation

Model Description

💻 Usage Examples

Basic Usage

pip install diffusers

from diffusers import UNet2DConditionModel, DiffusionPipeline, LCMScheduler
import torch

unet = UNet2DConditionModel.from_pretrained("briaai/BRIA-2.3-FAST", torch_dtype=torch.float16)
pipe = DiffusionPipeline.from_pretrained("briaai/BRIA-2.3-BETA", unet=unet, torch_dtype=torch.float16)
pipe.force_zeros_for_empty_prompt = False

pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
pipe.to("cuda")

prompt = "A portrait of a Beautiful and playful ethereal singer, golden designs, highly detailed, blurry background"

image = pipe(prompt, num_inference_steps=8, guidance_scale=1.0).images[0]

📄 License

Bria AI Model weights are open source for non - commercial use only, per the provided license.

💡 Usage Tip

⚠️ Important Note

Bria 3 generation is available: Bria 3.1

💡 Usage Tip

1. You must set pipe.force_zeros_for_empty_prompt = False.
2. Using negative prompt is recommended.
3. We support multiple aspect ratios, yet the resolution should overall consist of approximately 1024*1024 = 1M pixels, for example: (1024,1024), (1280, 768), (1344, 768), (832, 1216), (1152, 832), (1216, 832), (960,1088).
4. The Fast model works well with just 8 steps.
5. For the Fast models, use guidance_scale 1.0 or 0.0. Note that in this configuration, the negative prompt is not relevant.