FLUX.1 Kontext editing model now available as an Nvidia NIM Microservice

Image editing on your PC has typically been a complex task reserved for developers with technical knowledge. Nvidia is looking to change that, announcing the new FLUX.1 Kontext [dev] image editing model from Black Forest Labs is now available as an easy-to-deploy Nvidia NIM microservice for RTX AI PCs.

The FLUX.1 model is a generative tool designed to modify existing images using simple text and image prompts in a guided, step-by-step process. The magic here is the NIM (Nvidia Inference Microservice), which acts as a pre-packaged, highly optimised container. It takes the raw, complex AI model and handles all the difficult setup work that used to require expert knowledge, making it accessible to enthusiasts through popular tools like ComfyUI.

To make this feasible on consumer hardware, Nvidia and Black Forest Labs have heavily optimised the model. Through a process called quantisation (SVDQuant), the original 24 GB model has been shrunk down to a 12 GB FP8 version for GeForce RTX 40 Series GPUs and an even smaller 7 GB FP4 version for the latest RTX 50 Series cards. On top of this, the NIM uses Nvidia's TensorRT framework to deliver more than double the performance compared to running the original model in PyTorch.

For those interested in trying it out, getting started is straightforward. Start by installing the Nvidia AI Workbench and then get ComfyUI. Using the ComfyUI manager, install the NIM nodes for FLUX.1 and accept the model licenses on Black Forest Labs' FLUX.1 Kontext's [dev] Hugging Face. Once you click “Run”, the node will help download the required models and set up the appropriate process.

KitGuru says: Image editing models like these are a good example of simplification of complex processes using AI. Instead of losing hours in image edits, the FLUX.1 Kontext can make it probably better (especially if you're not well-versed in this area) and faster using simple prompts.

Become a Patron!