- Luma AI Inc. is advancing its initial concept, and as a crucial part of its strategy, it has established a robust cluster comprising over 3,000 Nvidia A100 graphics processing units.
- In the future, Luma will be directing its efforts toward assembling its Nvidia GPU cluster while expanding its team. The company aims to double its current 24-person workforce by the end of the coming year.
Luma AI Inc., a computer vision startup, announced the successful completion of a USD 43 million funding round. The company is set to utilize these funds to advance its mission of developing a new era of generative artificial intelligence models that can perceive and understand the surrounding world.
In the latest Series B funding round, premier venture capital firm Andreessen Horowitz took the lead, with support from notable contributors such as Amplify, Matrix, Nvidia Corp., and various undisclosed angel investors.
Originating as a straightforward smartphone app developer, Luma eventually gained traction with millions of users. The app ingeniously enabled individuals to leverage their smartphone cameras to generate 3D images of objects, eliminating the need for additional hardware.
The startup is advancing its initial concept, and as a crucial part of its strategy, it has established a robust cluster comprising over 3,000 Nvidia A100 graphics processing units. This formidable cluster will be dedicated to training cutting-edge AI models, aiming to achieve, in the words of Co-Founder and Chief Technology Officer Alex Yu, “see and understand, show and explain and eventually interact with the world.”
Before co-founding Luma, Yu was an AI researcher at the University of California, Berkeley. In parallel, Luma’s CEO Amit Jain, also a Co-Founder, contributed to Apple Inc., focusing on expanding the company’s Vision Pro multimedia experiences.
Luma’s ambition lies in crafting generative AI models capable of translating text descriptions into 3D objects. The inaugural model, Genie, was introduced in November, marking the initial step towards fulfilling this vision. Moving ahead, the startup aims to construct more sophisticated models endowed with advanced vision capabilities.
In the case of Genie, interactive 3D assets can be generated by users using text prompts such as “Gold ring with ruby gemstone.” According to the startup, the model has established a notable benchmark for swift text-to-3D generation, garnering thousands of users. Currently, Genie is being made widely accessible on the Luma smartphone app.
In a competitive landscape, the startup faces several rivals, such as Hypothetic, Auctoria AI Inc., and Kaedim Inc., each working towards developing similar capabilities. Established players like Stability AI Ltd. are also venturing into 3D model generation. At the same time, industry giants such as Autodesk Inc. and Nvidia have introduced applications like ClipForge, which is capable of generating 3D models from text descriptions, and Get3D, which transforms 2D images into 3D models.
In an interview, Yu conveyed that Luma’s 3D models distinguish themselves through superior fidelity. He elucidated that conventional models for 3D image generation are typically trained on two-dimensional images that often “mangle spaces, bodies, and movements.” It can be “very difficult to generate anything coherent and usable in the first few tries.”
In a blog post, Anjney Midha and Rajko Radovanovic, partners at Andreessen Horowitz, highlighted that Luma stands out in the crowd due to the exceptional abilities of its co-founders. They pointed out that Yu has previously authored groundbreaking papers on real-time neural rendering of 3D scenes and single-image generation. Simultaneously, Jain has gained popularity for his contributions to Apple Vision Pro.
They also referred to Luma’s Chief Scientist, Jiaming Song, who pioneered “diffusion models that have significantly improved state-of-the-art performance.”
In the future, Luma will be directing its efforts toward assembling its Nvidia GPU cluster while expanding its team. The company aims to double its current 24-person workforce by the end of the coming year.