Browse Models
Generate images and videos using a powerful low-level workflow graph builder - the fastest, most flexible, and most advanced visual generation UI.
Forge is a platform built on top of Stable Diffusion WebUI to make development easier, optimize resource management, speed up inference, and study experimental features.
Train your own LoRAs and finetunes for Stable Diffusion and Flux using this popular GUI for the Kohya trainers.
The FLUX.1 model family represents a significant advancement in text-to-image generation technology, developed by Black Forest Labs. Released in August 2024, this family consists of three distinct variants designed to address different use cases while sharing a common architectural foundation. The model family demonstrates exceptional capabilities in image synthesis, competing favorably with prominent closed-source alternatives such as Midjourney v6.0 and DALL·E 3.
All models in the FLUX.1 family share a sophisticated hybrid architecture that combines multimodal and parallel diffusion transformer blocks, scaled to 12 billion parameters. The architecture incorporates several advanced techniques, including flow matching, rotary positional embeddings, and parallel attention layers, which contribute to both performance and hardware efficiency. A notable technical innovation across the family is the use of guidance distillation during training, which helps maintain high-quality output while improving computational efficiency.
The FLUX.1 family includes three distinct variants, each optimized for specific use cases while maintaining the core architectural benefits of the family. The flagship variant, FLUX.1 pro, represents the highest-performing model in the family, though detailed specifications are not publicly available. The FLUX.1 dev serves as the open-weight version, designed for non-commercial applications and scientific research. The FLUX.1 schnell, released under the Apache 2.0 license, is specifically optimized for speed and local development, making it particularly suitable for personal use and rapid prototyping.
All models in the FLUX.1 family demonstrate remarkable versatility in image generation, supporting various aspect ratios and resolutions up to 2.0 megapixels. The family particularly excels in visual quality, prompt adherence, size/aspect variability, typography, and output diversity. A distinctive feature of the family is its ability to preserve output diversity from pretraining, ensuring a wide range of creative possibilities.
The FLUX.1 schnell variant particularly stands out for its ability to generate high-quality images in just 1-4 steps, making it a pioneer in few-step image generation at the time of its release. This achievement represents a significant advancement in making high-quality image generation more accessible and efficient for everyday users.
The FLUX.1 family is designed with broad accessibility in mind, offering multiple implementation options through various frameworks and platforms. The models can be accessed through the Hugging Face Diffusers library, which provides a standardized interface for model deployment. Additionally, local inference is supported through the ComfyUI platform, making the technology accessible to developers and enthusiasts working on local hardware.
While the FLUX.1 family demonstrates impressive capabilities, it shares common limitations across all variants. These include occasional failures in perfect prompt matching, with prompt following significantly influenced by prompting style. The models are not designed to provide factual information and may reflect existing societal biases. Additionally, all variants are subject to specific usage restrictions, particularly regarding the generation of harmful content or illegal materials.
The FLUX.1 family employs a tiered licensing approach that reflects the intended use cases of each variant. While FLUX.1 schnell is released under the permissive Apache 2.0 license, allowing for commercial use, FLUX.1 dev is restricted to non-commercial applications. All variants share common usage restrictions prohibiting the generation of harmful content, exploitation of minors, creation of disinformation, and other potentially harmful applications.
The release of the FLUX.1 family marks a significant milestone in democratizing access to advanced image generation technology. By offering variants optimized for different use cases and computing environments, the family addresses the needs of various user groups while maintaining high quality standards. The open-weight nature of certain variants and the implementation of efficient architectures suggest a commitment to advancing the field while promoting accessibility and responsible development practices.
Comprehensive documentation and resources are available for implementing and working with the FLUX.1 family. These include detailed integration guides through the Diffusers Documentation, reference implementations in the official GitHub repository, and specific usage guidelines outlined in the respective model licenses and acceptable use policies. These resources provide developers and researchers with the necessary tools and information to effectively utilize the models while adhering to ethical guidelines and technical best practices.