The Future of AI at Weasel Creative
Published April 2026
AI has become a bit like Marmite. People either love it or hate it. I don't sit in either camp, Marmite included, and I'm not caking it on. It's a tool, not a condiment. I just align myself with whatever is best for production, best for the brief, and best for getting the work done properly.
How we use AI in 3D production right now
Early concepting is where AI earns its keep. Testing styles, exploring visual directions, building rough mood references. It moves conversations forward with clients before committing production time to something that might not land.
It also works for non-hero elements. Background plates, environmental detail, supporting visuals that just need to add life to a scene without holding up under close inspection.
The Checkpoint Digital Product Passport film is the clearest example. Everything that mattered was built in Cinema 4D and Redshift. The background crowds were AI-generated. They didn't need to be perfect, just there. That saved weeks without affecting quality where it counted.
I also use AI for X-Particles Nexus scripts when I know what I'm building and just need to get there faster. And this entire website, every custom page, the SEO structure, all built with AI coding tools.
We still work in Cinema 4D, Redshift, and X-Particles because of the creative control they give us. That's a production decision. AI just can't do those things yet. It doesn't understand the world volumetrically and leans heavily on its training data. For product accuracy, lighting control, and consistent animation, that's not good enough.
What's next: AI style transfer with 3D control
The idea is to use greyscale renders from Cinema 4D as the foundation, controlling movement, camera, timing, and composition entirely in 3D, then applying AI style transfer on top. Quicker cinematic looks and bigger storytelling without the production overhead that would normally come with it.
The risk is frame-to-frame consistency. If the style drifts or flickers, it breaks. If it can't hold over 150+ frames, we'd take an AI-first approach instead and be upfront about that. It could be cheaper, but that's where the testing comes in. The time spent getting reliable results, the token spend, the iterations to get it right, all of that might offset the savings. Until the workflow is proven and repeatable, the cost is an unknown.
I've played with open source Stable Diffusion models in ComfyUI. Out of the box, results weren't great. Poor consistency, broken outputs, and the occasional four-headed nightmare fuel when you change the aspect ratio. That said, with ControlNet and proper node setups, the results improve significantly. That's the gap to close.
I'm planning to build an AI reel around this, but it needs to earn its place first. Promising is not the same as proven.
The bottom line
At the end of the day, it's about tried and tested workflows I can fully get behind. I don't like getting backed into a corner with work I can't fix, and that's the risk with relying too heavily on tools that might not be around in one to five years. The goal is to produce work, creatively and production-wise, that would normally need a team, from a team of one. But that only works if the foundation is solid and every tool in the chain is something I trust.
Build properly. Use AI where it genuinely helps. Keep testing.
AI in 3D animation: common questions
Is AI replacing 3D animation? No. AI supports parts of the workflow, but it can't replace the control, accuracy, and consistency that commercial CGI needs.
When should you use AI vs CGI? AI for concepting, non-hero elements, and coding. CGI for anything product-accurate, consistent, or commercially reliable. For a deeper breakdown, read AI vs CGI for Product Animation.
What tools work best for AI in 3D animation? ChatGPT and Claude for coding. ComfyUI with Stable Diffusion and ControlNet for style transfer. Nano Banana 2 for concepting.
If you want to test where AI actually adds value in your workflow, get in touch.