Sponsored by
Get generative AI without cloud dependency thanks to ComfyUI
If you’ve never heard of ComfyUI, then let me initiate you. ComfyUI is a node-based workflow tool for running local AI image and video generation models on NVIDIA RTX hardware. If the idea of running AI locally on your own machine is a new concept, then this article will help you understand not only the benefits of doing so but also some of the workflows.
Let’s begin with a couple of the major benefits, with the first being that you can ditch cloud dependency. Once you’ve downloaded a model into ComfyUI, it’s then yours to use without needing the software's servers to be behaving themselves. Better still, you can create images and videos without needing any internet connection at all.
The other significant benefit is related to privacy and security. If you’re relying on cloud AI, then your data could be at risk, but if everything is run locally, then you bypass having to upload and download any data. That means that if you have a sensitive source image that you don’t want to share with the cloud, then you don’t have to with ComfyUI.
Local image generation with Flux.2
With RTX acceleration, ComfyUI can run the Flux.2 model, which is perfect for rapid image generation. To get started, all you have to do is download ComfyUI and install it locally. Then it’s just a case of downloading the Flux.2 model. It’s a pretty big file, so make sure you have the gigabytes to accommodate it!
From there it’s possible to generate photoreal images, enter natural prompts, and blend up to 10 references for consistent results. One of the best use cases that I’ve seen for this model is being able to place a product within any context. Flux.2 Ensures that the lighting adapts and that the physics work. Flux says that it "looks real because the model understands real".
Flux.2 is also very good at keeping the scene structure so that key subjects “stay perfectly placed while the rest of the picture evolves as you wish". This makes it possible to perform rapid style swaps and precise adjustments without worrying that the contents of your image are going to completely regenerate every single time. In short, you’re in full creative control.
Motion concepts with LTX Video 2.3
Flux.2 isn’t the only model that ComfyUI supports, far from it. If you’re looking to generate dynamic video sequences from images, then I would turn to the LTX Video 2.3 model, which offers fantastic motion consistency and fine details.
The LTX-2.3 model from Lightricks looks to prioritise the integrity of fine details within a source image while translating that into a moving image. It’s no small feat, but LTX Video does a great job.
In terms of the specific workflow, you’ll need to utilise the LoadImage node for importing the source image and the SaveVideo node for exporting the final video output. In between you’ll require the beautifully named(!) b94257db-cdc1-45d3-8913-ca61e782d9c1 node for processing and enhancing audio quality.
This specific workflow is a game-changer for creatives wanting to generate a small amount of dynamic movement to an image. It could be bringing some personality to a portrait photo or creating b-roll when only photos were captured on the shoot.
ComfyUI and RTX technology is the perfect partnership
ComfyUI is the perfect practical prototyping tool for creatives who want generative AI without cloud dependency. The wide range of models that are supported means that artists can carry out various different workflows within one familiar software package.
If you’d like the very best performance from ComfyUI, then pairing it with an NVIDIA RTX GPU is a must. NVIDIA states that “performance is 2.5x faster and VRAM is reduced by 60% with NVIDIA GeForce RTX 50 Series GPUs’ NVFP4 format, and performance is 1.7x faster and VRAM is reduced by 40% with FP8.”
AI image and video generation is traditionally a very time-consuming process, but with RTX technology you can deliver when facing quick turnarounds and tight deadlines.
