A few years back, Alex Yu and Amit Jain joined forces to establish a company enabling individuals to capture objects in 3D using their smartphones, without the need for additional equipment. Yu was an AI researcher at UC Berkeley, and Jain was an Apple employee working on Vision Pro’s multimedia experiences.
Their company, Luma, introduced a smartphone app in 2021, which quickly gained traction and amassed over two million users. Now, as generative AI technology becomes more prevalent, Yu and Jain aim to expand Luma into something greater than they had initially envisioned.
Luma has announced plans to utilize a compute cluster of approximately 3,000 Nvidia A100 GPUs to train new AI models with the capability to “see and understand, show and explain and eventually interact with [the] world,” as described by Yu.
The first phase of this initiative involves creating models capable of generating 3D objects from text descriptions. Luma has already launched one model, called Genie, on its Discord server. The next phase will focus on developing “next gen” generative AI models to address the shortcomings of current-gen GenAI, according to Yu.
To realize this vision, Luma has secured $43 million in a Series B round, with Andreessen Horowitz among the backers. This investment values Luma at between $200 million and $300 million, bringing its total funding to over $70 million.
Luma’s current primary focus, creating 3D model-generating AI models, faces growing competition from platforms like 3DFY and Scenario, as well as startups including Hypothetic, Kaedim, Auctoria, and Mirage. Additionally, established players like Autodesk and Nvidia are also venturing into this sector with apps like Get3D and ClipForge.
Image Credits: Luma
Luma aims to differentiate its tools by focusing on fidelity, according to Yu. He emphasizes that current models, trained on two-dimensional images, often struggle to generate coherent and usable scenes, which Luma aims to improve with its advanced generative photorealistic technologies.
While the improved version of Genie has been launched, Luma is continuously working on developing more capable generative AI models. The company plans to double its 24-person workforce by the end of next year and establish a model-running server cluster of “thousands” of GPUs.
Image Credits: Luma
Jain expressed the company’s commitment to growing the team and accelerating progress to build vastly more capable, intelligent, and useful visual models for users, following the recent round of funding.
