flux

Official inference repo for FLUX.1 models

open-sourceagent-frameworks
25.4k
Stars
+113
Stars/month
0
Releases (6m)

Star Growth

+14 (0.1%)
24.8k25.4k25.9kMar 27Apr 1

Overview

FLUX is Black Forest Labs' official inference repository for running image generation and editing with their FLUX.1 open-weight models. The toolkit provides minimal inference code for multiple specialized models including text-to-image generation, inpainting/outpainting, and structural conditioning based on edge maps or depth information. With over 25,000 GitHub stars, FLUX offers a comprehensive suite of models ranging from the Apache 2.0 licensed FLUX.1 [schnell] for commercial use to more advanced [dev] variants for research and non-commercial applications. The repository supports both standard PyTorch installations and optimized TensorRT deployments for enhanced performance on NVIDIA hardware. FLUX represents a significant advancement in open-source image generation, providing researchers and developers with state-of-the-art capabilities that were previously only available through proprietary APIs.

Deep Analysis

Capabilities

  • Open-source image generation and editing toolkit with multiple diffusion model variants
  • Text-to-image generation with high-quality outputs
  • In/out-painting and image editing workflows
  • Structural conditioning via Canny edge detection and depth mapping
  • Image variation generation and style transfer
  • Local inference support with optional TensorRT GPU acceleration
  • Multiple specialized models: Kontext, Redux, Krea, schnell, dev

🔗 Integrations

PyTorchHugging FaceTensorRTBlack Forest Labs API

Best For

  • Developers and researchers needing state-of-the-art open-weight image generation
  • Commercial enterprises requiring licensed, self-hosted image generation
  • Creative professionals using programmatic image generation pipelines

Languages

Python

Deployment

Local inference (Python 3.10+)Cloud API via docs.bfl.aiHugging Face model download

Pricing Detail

Free: Open-weight models: schnell (Apache 2.0), dev (non-commercial)
Paid: Commercial licensing at bfl.ai/pricing; API usage-based pricing

Known Limitations

  • Large models require significant GPU VRAM
  • Dev variants restricted to non-commercial use
  • Commercial licensing requires separate agreement and monthly fees
  • Python 3.10+ required

Pros

  • + Multiple specialized models for different image generation tasks including text-to-image, inpainting, and structural conditioning
  • + Open-weight architecture with both commercial (schnell) and research (dev) licensing options available
  • + TensorRT optimization support for high-performance inference on NVIDIA hardware

Cons

  • - Most advanced models (dev variants) are restricted to non-commercial use only
  • - Requires substantial computational resources and GPU memory for optimal performance
  • - Limited to inference only - no training code or fine-tuning capabilities included

Use Cases

  • Creating high-quality images from text prompts for commercial or research projects
  • Performing inpainting and outpainting to edit or extend existing images
  • Generating images with structural conditioning using edge maps or depth information

Getting Started

Clone the repository from GitHub and navigate to the flux directory. Create a Python 3.10 virtual environment and activate it. Install the package with dependencies using pip install -e ".[all]" and refer to the docs for specific model usage instructions.

Compare flux