Black Forest Labs Introduces AI-Powered Flux.1 Tools for Advanced Image Editing

Black Forest Labs launched 4 new synthetic intelligence (AI) instruments for its base text-to-image Flux.1 AI mannequin final week. These 4 AI instruments are based mostly on separate fashions designed to execute particular picture modifying duties throughout the picture generator. The firm claims that these instruments will provide granular management over the output photographs and let customers protect the important thing parts whereas experimenting with completely different kinds. The modifying instruments can be found individually as developer fashions in open entry and professional fashions.

In a blog post, the AI agency detailed the 4 new picture modifying instruments for Flux.1 AI mannequin. Developers can open-access the 4 instruments in separate fashions throughout the dev mannequin sequence, whereas customers will get entry to the complete model through BFL API.

The Flux.1 Fill is an inpainting and outpainting device that may edit the small print inside a picture or increase the boundaries of a picture utilizing textual content prompts and a binary masks. Based on inside testing, the corporate claimed that the professional model of the device outperforms competing instruments comparable to Ideogram 2.0. The developer model of the device is out there beneath the Flux Dev License and may be found on Hugging Face and GitHub. The professional model may be accessed through the BFL API.

Flux.1 Depth and Flux.1 Canny instruments let customers carry out structural conditioning of output throughout picture transformations. The Depth device preserves the generated picture’s construction via depth maps and retains it intact whereas customers make a text-guide edit. Similarly, the Canny device preserves the construction by accessing the output’s canny edges. These are useful throughout retexturing-based edits.

The firm claimed that the instruments outperform comparable instruments supplied by rivals comparable to Midjourney and InstantX. The full model provides most efficiency whereas the Low-rank adaptation (LoRA) model for builders permits for simpler deployment. It may be discovered here.

Finally, the Flux.1 Redux permits customers to generate picture variations based mostly on an enter picture. Black Forest Labs claims the device can reproduce the picture with slight variation, which may be refined additional. It additionally permits picture restyling through prompts. Also, this device is supported by the Flux1.1 [pro] Ultra, the corporate’s flagship picture era mannequin. The mannequin weights may be discovered here.

All of the AI fashions may even be accessible through third-party platforms comparable to Fal.ai, Replicate, Together.ai, Freepik, and Krea.ai.