Developed by former Stability AI engineers, FLUX.1 has generated a lot of excitement online for its photorealistic capabilities.
The latest series of artificial intelligence (AI)-generated images to go viral on social media show close-ups of people speaking into microphones, most likely on stage at a conference or other event. The lifelike rendering of the skin, hair, and wrinkles of the men and women in the photos, which first appeared on a Reddit post dedicated to AI-generated content, has sparked much discussion online.
These images were generated by Black Forest Labs, a German startup, using their recently released FLUX.1 text-to-image artificial intelligence algorithm.
Although similar images have appeared online before, those generated by FLUX.1 show that these AI models can now depict fictional individuals with nothing strange about them (other than the text on their badges). Because FLUX.1 is an open-source AI model, it serves as a sign that technology is becoming both more complex and more accessible.
However, it should be emphasized that the AI-generated images generated by FLUX.1, which have received widespread attention, have been further improved by applying the Low-Rank Adaptation (LoRA) technique, which is designed to improve and optimize the results of large language models.
What exactly is FLUX.1?
Three different versions of FLUX.1, which was released on August 1, are available to users. The high-end version for commercial use is the AI model called “Pro”. The faster version of FLUX.1 with open weights is known as “Schnell,” which translates to “quick” or “fast” in German. Meanwhile, the “Dev” version of FLUX.1 includes open weights and is designed for non-commercial use.
A combination of diffusion and transformer approaches are purportedly used in the architecture of the 12 billion parameter AI model. The parameter sizes of the Stability Diffusion 3 model suite, for example, range from 800 million to 8 billion. Using training techniques like flow matching, Black Forest Labs announced in a blog post that FLUX.1 has been further optimized. The source of the data used to train the AI model was not disclosed by the corporation, though.
Black Forest Labs, an AI startup founded by a group of AI researchers and engineers, some of whom previously held key positions at Stability AI, is the creator of FLUX.1. In fact, Dominik Lorenz, Andreas Blattmann, and Robin Rombach were reportedly employed by the AI company before the release of its Stability Diffusion 3 (SD3) model.
According to a report by ArsTechnica, Black Forest Labs recently raised over $31 million in a seed funding round that included investments from Silicon Valley investors such as Andreessen Horowitz, General Catalyst, and MätchVC.AI researcher Matthias Bethge, as well as former Disney president Michael Ovitz, who also joined as an advisor.
What does FLUX.1 do?
It seems that FLUX.1 can create human hands and legs. Due to shortcomings in the training datasets, the images produced by the AI models had problems accurately depicting human limbs.
According to the company, “a wide range of aspect ratios and resolutions in 0.1 and 2.0 megapixels are supported by all FLUX.1 model variants.” Going forward, Black Forest Labs said it is developing a text-to-video generator to compete with Kuaishou’s Kling, Runway’s Gen-3 Alpha, and OpenAI‘s Sora.
“Our video models will enable accurate creation and editing at unprecedented speed and high definition.”
Hugging Face and other AI developer platforms make the weights of FLUX.1 [dev] and FLUX.1 [schnell] publicly available. Replicate and Fal, two AI cloud hosting platforms, offer these models for direct use. Freepik, a stock photo website, has added FLUX.1 to its AI toolkit.
