Last week, Black Forest Labs unveiled four new artificial intelligence (AI) tools designed to enhance its Flux.1 text-to-image model. These tools, each built on distinct models, are specifically engineered for various image editing tasks. The company asserts that these innovations will provide users with detailed control over their output images, enabling them to maintain essential elements while exploring different stylistic approaches. Available in both open access and pro models, the editing tools can be utilized separately.
In a recent blog post, Black Forest Labs elaborated on these four image editing tools. Developers can access these tools through individual models within the development model series, while users can acquire the full version through the BFL API.
Today, we are excited to release FLUX.1 Tools, a suite of models designed to add control and steerability to our base text-to-image model FLUX.1, enabling the modification and re-creation of real and generated images. Learn more in our blogpost: https://t.co/J5Bc8fVGEc pic.twitter.com/7lEl74XYV4
— Black Forest Labs (@bfl_ml) November 21, 2024
The Flux.1 Fill tool enables both inpainting and outpainting, allowing users to modify image details or extend the edges of an image using textual prompts and a binary mask. According to internal tests, the pro version reportedly outperforms rival tools such as Ideogram 2.0. The developer version is available under the Flux Dev License and can be accessed on Hugging Face and GitHub, while the pro version is accessible through the BFL API.
Flux.1 Depth and Flux.1 Canny tools facilitate structural conditioning during image transformations. The Depth tool ensures the integrity of the generated image’s structure using depth maps, enabling users to make text-guided edits without compromising the image’s form. The Canny tool similarly maintains structure by leveraging the output’s canny edges, which is beneficial for retexturing edits.
Black Forest Labs claims that its tools surpass similar offerings from competitors like Midjourney and InstantX. While the full version delivers optimal performance, a Low-rank adaptation (LoRA) version designed for developers simplifies deployment. This version can be located here.
Lastly, Flux.1 Redux allows users to create variations based on a selected image. The tool can replicate an image with minor alterations, offering further refinements, and allows for image restyling through prompts. This tool is supported by the Flux1.1 [pro] Ultra, which is the company’s premier image generation model. Model weights are available here.
All AI models introduced by Black Forest Labs will also be accessible on third-party platforms, including Fal.ai, Replicate, Together.ai, Freepik, and Krea.ai.
