Flux

We release the FLUX.1 suite of text-to-image models that define a new state-of-the-art in image detail, prompt adherence, style diversity and scene complexity for text-to-image synthesis. To strike a balance between accessibility and model capabilities, FLUX.1 comes in three variants: FLUX.1 [pro], FLUX.1 [dev] and FLUX.1 [schnell]: All public FLUX.1 models are based on a hybrid architecture of multimodal and parallel diffusion transformer blocks and scaled to 12B parameters. We improve over previous state-of-the-art diffusion models by building on flow matching , a general and conceptually simple method for training generative models, which includes diffusion as a special case. In addition, we increase model performance and improve hardware efficiency by incorporating rotary positional embeddings and parallel attention layers . We will publish a more detailed tech report in the near future. FLUX.1 defines the new state-of-the-art in image synthesis. Our models set new standards in their respective model class. FLUX.1 [pro] and [dev] surpass popular models like Midjourney v6.0, DALL·E 3 (HD) and SD3-Ultra in each of the following aspects: Visual Quality, Prompt Following, Size/Aspect Variability, Typography and Output Diversity. FLUX.1 [schnell] is the most advanced few-step model to date, outperforming not even its in-class competitors but also strong non-distilled models like Midjourney v6.0 and DALL·E 3 (HD) . Our models are specifically finetuned to preserve the entire output diversity from pretraining. Compared to the current state-of-the-art they offer drastically improved possibilities as shown below Today we release the FLUX.1 text-to-image model suite. With their strong creative capabilities, these models serve as a powerful foundation for our upcoming suite of competitive generative text-to-video systems. Our video models will unlock precise creation and editing at high definition and unprecedented speed. We are committed to continue pioneering the future of generative media. We are hiring exceptionally strong machine learning and backend engineers. If you are interested in joining our team, reach out to [email protected].

This repo contains minimal inference code to run image generation & editing with our Flux open-weight models. If you would like to install the repository with TensorRT support, you currently need to install a PyTorch image from NVIDIA instead. First install enroot, next follow the steps below: We are offering an extensive suite of open-weight models. For more information about the individual models, please refer to the link under Usage . FLUX.1 [schnell] Text to Image apache-2.0 FLUX.1 [dev] Text to Image FLUX.1-dev Non-Commercial License FLUX.1 Fill [dev] In/Out-painting FLUX.1-dev Non-Commercial License FLUX.1 Canny [dev] Structural Conditioning FLUX.1-dev Non-Commercial License FLUX.1 Depth [dev] Structural Conditioning FLUX.1-dev Non-Commercial License FLUX.1 Canny [dev] LoRA Structural Conditioning FLUX.1-dev Non-Commercial License FLUX.1 Depth [dev] LoRA Structural Conditioning FLUX.1-dev Non-Commercial License FLUX.1 Redux [dev] Image variation FLUX.1-dev Non-Commercial License FLUX.1 Kontext [dev] Image editing FLUX.1-dev Non-Commercial License FLUX.1 Krea [dev] Text to Image FLUX.1-dev Non-Commercial License The weights of the autoencoder are also released under apache-2.0 and can be found in the HuggingFace repos above. Our API offers access to all models including our Pro tier non-open weight models. Check out our API documentation docs.bfl.ai to learn more. As the fee is based on a monthly usage, we provide code to automatically track your usage via the BFL API. To enable usage tracking please select track usage in the cli or click the corresponding checkmark in our provided demos. We provide a reference implementation for running FLUX.1 with usage tracking enabled for commercial licensing. This can be customized as needed as long as the usage reporting is accurate. The above reporting logic works similarly for FLUX.1 [dev] and FLUX.1 Tools [dev]. Note that this is only required when using one or more of our open weights models commercially. More information on the commercial licensing can be found at the BFL Helpdesk.