Nvidia Shows Off ‘Magic3D’ Tech That Allows You To Print Models Just From Text

Aside from text-to-image generators and chatbots, there are other forms of AI technology on the market that are just as, if not more, impressive. Nvidia’s ‘Magic3D’, demonstrated in November, allows users to generate full 3D-printed models just from a simple text prompt. 

According to the developers, the tool produces 3D mesh models with “unmatched quality,” allowing users to discover new ways to make use of 3D synthesis, image conditioning techniques, and prompt-based editing methods. 

More importantly, this technology could help democratize 3D printing technology as a whole, with researchers stating in a paper that the novel generator would allow anyone to produce their own 3D models even without specialized knowledge. 

Furthermore, Magic3D—which takes a low-resolution model and turns it into one with a higher resolution—could speed up the development of video games, virtual reality, and even feature in special effects for film and television. 

The two-stage system functions similarly to that of 2D generator DreamFusion, with the 3D iteration said to create models twice as quickly as its predecessor. A coarse-to-fine optimization system first produces a low-resolution diffusion, before turning the prompt into an actual model.

With this incredible technology, you’ll now be able to input a description of the idea you wish to create, sit back for an estimated 40 minutes, and poof—you’ve got yourself your very own 3D-printed model. 

[via Interesting Engineering and Nvidia, cover image via Nvidia]