🚨 All generations are PUBLIC.

The purpose of this site is to allow everyone to openly experiment with Flux Pro, so that we can assess it's strengths and weaknesses collectively rather than in isolation, and to learn about it's prompting styles from one another.

What is Flux?

Flux is a new state-of-the-art open source text-to-image model developed by Black Forest Labs, the original team behind Stable Diffusion. With an impressive 12 billion parameters, Flux delivers high-quality images reminiscent of Midjourney's aesthetics.

Key Features

  • Hybrid architecture: Flux uses a hybrid architecture with multimodal and parallel diffusion transformer blocks.
  • Diverse aspect ratios and resolutions: The model supports various aspect ratios and resolutions up to 2.0 megapixels.
  • Strong prompt following: Flux demonstrates solid prompt following capabilities, able to accurately depict elements like hands, feet, and specific styles.

Model Variations

Flux comes in three powerful variations:

  1. FLUX.1 [dev]: The base 12B parameter model, open-sourced with a non-commercial license for the community to build upon.
  2. FLUX.1 [schnell]: A distilled version of the base model that operates up to 10 times faster. Licensed under Apache 2.
  3. FLUX.1 [pro]: A closed-source version only available through API.

Performance

Early reviews indicate Flux delivers what many hoped for from Stable Diffusion 3 - high-quality, aesthetically pleasing images with strong coherence to prompts. It can run on GPUs with 12GB VRAM using lower precision settings.

Black Forest Labs, founded by Stable Diffusion creators Robin Rombach and Patrick Esser, secured $31 million in seed funding to support Flux's development. The team plans to release text-to-video models in the future as well.

The release of Flux marks a major step forward in making high-quality, open source text-to-image generation more widely accessible to creators and developers. You can now even train your own lora!

What is this site for?

Evaluating the quality of a new model is challenging, given the variations in seeds, steps, schedulers, and only cherry-picked examples making the headlines. To address this, we are establishing a community gallery to generate Flux Pro images at no cost, aiming to at least increase the sample size. All generations will be public. Please note that this initiative is independent and not affiliated with Black Forest Labs.