During the IFA 2025 conference in Berlin, Nvidia unveiled a suite of improvements to its generative AI tools for 3D artists and developers. The two main things the company has introduced are a new AI Blueprint for 3D Object Generation, alongside the Microsoft TRELLIS NIM microservice, to make scene prototyping significantly faster and easier. The pitch is that these tools will cut down on manual work and give creators more time to focus on the artistic side of design, but is this the full story?

The new 3D Object Generation Blueprint uses Llama (the Large Language Model) and NVIDIA SANA to enable users to produce up to 20 prototype objects directly from a text prompt. These objects can be refined, modified, and turned into high-quality models in a fraction of the time it would normally take.

Microsoft TRELLIS, a research-driven model built to generate detailed 3D assets from text or images, handling intricate geometry and textures. Integrated as an NVIDIA NIM microservice, TRELLIS benefits from GPU acceleration on RTX cards, running 20% faster thanks to PyTorch optimizations. On an RTX 5090, the improvement translates to roughly six seconds saved per generated object. For artists working with large numbers of assets, this means noticeable gains in speed and efficiency.

How much will this benefit 3D artists?

From the perspective of most 3D artists at least, AI has so far been a net negative, resulting in lay-offs and, for those that do retain their jobs, often being forced to base their designs on AI-generated concepts handed to them by their employers, reducing any creativity (Kamikaze Shortbus gives a good summary from the perspective of someone who worked in the industry).

Businesses will naturally try and use every tool available to them to improve their bottom line, but this will naturally also lead to fewer professional artists and less human-generated art being made (could you even call imagery generated by a non-sentient being art? Discuss).

‘Prototyping’ is broad enough a term that it includes the genuinely tedious, non-creative parts of the initial design stages of software development or 3D art (which could be a good thing, like the many time-saving tools we currently use), but also the early stages of creative genesis with something machine-generated.

Looking at Nvidia’s latest blog post, which goes into these features in detail, one can’t help but think it’s probably both: using NVIDIA SANA, the user can ‘start by providing an artistic idea through a prompt, and the blueprint's built-in large language model (LLM) will brainstorm 20 possible objects to include in the scene’.

About the Author

Aaron Ritchie

Aaron's laptop knowledge makes him the go-to guy on PC Guide. But he still finds time for features, deals and much more.