Can a local AI assistant improve your creative process?

I’ve not had much experience of using local LLMs, but I’m coming across them in my life with increasing frequency. Local LLMs are ideal as collaborators for image and video generation, ideation, copywriting direction, and client communication.

Being able to take ideas from concept to client-ready output without leaving the desktop is a big win for convenience and privacy. I had previously discounted myself from using local LLMs but, in my limited experience, have found them surprisingly easy to use.

Below we highlight two of the best locally hosted engines for embracing AI and transforming your creative processes. And, if you’re looking for more ideas, then check out how I use AI in my creative workflow.

1. ComfyUI

Before this week I had never come across ComfyUI, but now I love it. For those unaware of the tool, ComfyUI is an AI creation engine specifically tailored towards visual professionals who demand control over every model, every parameter, and every output.

I installed ComfyUI and immediately began utilising the node-based system, using a pre-provided template as a starting point. The nodes provide the interface for interacting with settings and parameters, and it is the collection of nodes which are used to generate an output.

After install, I was able to generate ideas as well as finished images within 10 to 15 minutes, although it took longer to become more familiar with the interface and what is possible beyond the basics. I imagine I’ll be using ComfyUI more often in the future and can see it being particularly useful for rapid prototyping.

If the node system is a little overwhelming, there's also a new App View which presents workflows in a simplified interface. In this view, users only need to enter a prompt, adjust simple parameters and hit generate.

To take full advantage of ComfyUI, you’ll need an RTX card. According to NVIDIA, “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.” Basically, if you want the fastest processing, it has to be RTX.

2. LTX Video 2.3

AI video doesn’t get much better than with LTX Video 2.3, especially if you own an NVIDIA RTX card. With a wide range of next-level tools, users can create varied content for a broad selection of end uses.

Take the Audio to Video feature, which enables users to generate video using voice, music, and sound effects as a source for defining structure, pacing, and motion. This is an ideal solution for production-grade workflows, including avatars, podcasts, and voice-driven clips.

If it’s creating social media content that you’re interested in, then LTX Video has you covered there as well. With Native Portrait, it’s possible to generate vertical video up to 1080 x 1920 pixels. And this isn’t just cropping landscape footage; the model is specifically trained on portrait-orientation data so you know it’s going to deliver the best results.

LTX Video really comes into its own when you begin customising the model with your own data and identity. By doing so, the model adapts to your worlds, characters, and creative DNA. That means you can generate AI video content that is informed by your specific design choices rather than whatever the default model dictates.

Embracing local LLMs

Locally hosted LLMs might sound complex and complicated, but it’s a lot easier than it sounds, especially with the tools we’ve explored above. If you own an RTX GPU already, then giving them a go is a no-brainer. Or, if you’re planning on integrating local LLMs into your workflow, then buying the best that NVIDIA has to offer is a must.