Enlarge / AI-generated visualization crafted by FLUX.1 development team: “An enchanting universal queen raising her hands, with a face depicted in the background.” (credit: FLUX.1)
A significant announcement was made on Thursday by Black Forest Labs, an innovative startup specializing in artificial intelligence. The company introduced its first range of text-to-image AI models under the designation FLUX.1. Based in Germany, this organization was established by a group of researchers who played key roles in creating the technology behind Stable Diffusion and were instrumental in developing the latent diffusion methodology. Their mission is to push boundaries by delivering sophisticated generative AI solutions for images and videos.
The unveiling of FLUX.1 follows closely behind Stability AI’s controversial launch of Stable Diffusion 3 Medium earlier this June—a release that attracted harsh reviews from enthusiasts within the image-synthesis community due to its inadequate representation of human anatomy, leading to rampant sharing of distorted limbs and bodies across various social networks. This unfortunate rollout occurred shortly after three prominent engineers—Robin Rombach, Andreas Blattmann, and Dominik Lorenz—left Stability AI to co-establish Black Forest Labs alongside latent diffusion collaborator Patrick Esser and several other talented individuals.
Black Forest Labs has launched its services with three distinct variants of the FLUX.1 text-to-image models; these include a premium commercial “pro” edition, a mid-tier “dev” version featuring open weights geared towards non-commercial applications, and an expedited open-weights edition referred to as “schnell,” translating to quick or fast from German terminology. The company asserts that their models deliver superior performance when compared with existing alternatives such as Midjourney and DALL-E, particularly excelling in image quality as well as fidelity to provided textual prompts.
Read 9 remaining paragraphs | Comments
