Luma AI’s Genie lets anyone make 3D objects from text

Luma AI, a startup that specializes in creating 3D content from video and text, has unveiled a new tool that allows anyone to generate realistic 3D models in seconds.

The tool, called Genie, is a research preview of an all-new kind of generative 3D foundation model that can create 3D things from natural language prompts.

Genie is currently available for free on Discord, where users can join a channel and type in what they want to create. Genie will then generate four different 3D models that match the description, along with a link to download them in GLB format. Users can also customize the materials and styles of the models, as well as view them in augmented reality.

The tool is powered by a deep neural network that has been trained on a large dataset of 3D shapes, textures, and scenes. The network can learn the semantic and geometric relationships between words and 3D objects, as well as synthesize novel shapes that are consistent with the input. Genie can handle a wide range of domains, from furniture and animals to vehicles and buildings.

The vision and impact of Luma AI

Genie is a potential game-changer for the art world. According to Luma AI’s co-founder and CEO Amit Patel, Genie is part of the company’s vision to democratize 3D content creation and enable anyone to make videos, scenes, and worlds that look plausible and are useful.

“The core belief at Luma is that all visual generative models need to reason and work in 3D to make videos, scenes, and worlds that look plausible and are useful,” Patel tweeted. “Today we are sharing Genie, a tiny peek into our foundation model research to show what’s possible. It’s really fast at making up things and a good bit of fun! And it will get better even faster!”

Feedback and reaction from the community

Genie has already attracted a lot of attention and praise from the online community. Many users have shared their creations on Twitter using the hashtag #MadeWithGenie. Some users have also imported their models into other software such as Blender or Unity to further edit or animate them.

“Pretty compelling Text-to-3D. Prompt was ‘modern purple sofa’. Generated in 14 secs (with 3 others) and the GLB imports into Blender in another 5 seconds,” tweeted Andrew Price, a popular Blender artist and instructor.

“I’m blown away by the new @LumaLabsAI tool. ‘Genie’ created these crocos for me in just 20 seconds??,” tweeted Bruno Pontillo, a digital artist and game developer.