The "FLUX online version" has added a series of important features, including ControlNet, LoRA, and other modules, which are ready to use without manual configuration. Users can even train their own LoRA models online and share them.
After the launch of FLUX, created by the original Stable Diffusion team, fal.ai quickly followed up and integrated new features such as ControlNet and LoRA. The out-of-the-box usability and one-click sharing advantages of fal have enabled users to create various creative works.
Fal is one of the two official online inference platforms partnered with the FLUX development team. Jonathan Fischoff, a member of the fal team, has been extensively reposting user works on social media and mentioned that they are developing features combining ControlNet and LoRA with image-to-image generation.
Users have created various styles of images using FLUX+LoRA+ControlNet, such as cyberpunk-style bedrooms and Chinese-style scenes. ControlNet can achieve effects like outline control and pose control. The LoRA module can modify characters into different styles like paper-cutting and painting.
The fal platform currently has 5 FLUX-related models, with the latest version supporting ControlNet and LoRA. Basic functions include text-to-image, image-to-image, and image inpainting. The ControlNet section uses a collection developed by the InstantX team, supporting 7 control modes. The LoRA function supports multiple module stacking and allows online training and sharing of custom modules.
Using the fal platform requires payment, at $0.075 per million pixels. New users are given a $1 balance. FLUX itself is free and open-source, allowing for self-deployment.