What Luma AI does
Luma AI (lumalabs.ai) builds multimodal “world model” systems that target next-step creative and physical-world intelligence, combining generative video with 3D capture/reconstruction and agentic workflows. The company’s consumer-facing product suite is organized around Dream Machine (video generation and editing) and adjacent capabilities such as 3D generative modeling (Genie) and interactive 3D scene capture (Interactive Scenes), with developer access via APIs and an agent layer intended to coordinate multi-step generation tasks. Luma also sells enterprise offerings aimed at integrating visual creation into brand and production workflows, positioning its models and creative tools as an end-to-end creative partner for teams.
In terms of technical differentiation, Luma emphasizes multimodal foundation modeling spanning video, 3D, and (in its roadmap language) unified understanding and generation. For example, the company describes Dream Machine as its generative ecosystem and frames recent releases as modular video-generation and camera/control concepts (e.g., Ray-family and composable camera motion). In the 3D direction, Luma describes Interactive Scenes and Genie as building blocks for generating and capturing viewable 3D assets and “worlds” that can be edited, rendered, and used downstream.
Strategically, Luma appears to be scaling both product reach and compute for training/deployment. In 2025 the company announced a $900M Series C led by HUMAIN, with AMD Ventures participation and continued involvement from prior investors including Andreessen Horowitz (a16z), Amplify Partners, and Matrix Partners. In parallel, Luma’s public materials indicate it is expanding engineering and research teams and pushing new video model releases and APIs (e.g., video generation and modify-video-style functionality), supported by its stated customer/partner adoption across media, entertainment, and advertising.