Luma AI

Luma AI is an American generative artificial intelligence company that builds multimodal models for video, image and three-dimensional content. Founded in 2021 as a pioneer of neural-rendering 3D capture, it rose to prominence with the Dream Machine video generator in 2024 and the Ray family of video models, which the company describes as the first video models designed to reason about a prompt before generating footage.^[1][2]

History

Luma AI was founded in September 2021 by Amit Jain, a former Apple engineer who worked on the passthrough feature of the Apple Vision Pro and on integrating the first LiDAR sensors into the iPhone, together with Alex Yu and Alberto Taiuti. Its earliest products centred on neural radiance fields, a technique for capturing real-world objects and scenes in three dimensions; a free iPhone capture app served both as a product and as a way to gather training data, followed by Genie, a text-to-3D model that attracted tens of thousands of users.^[7]

The consumer breakthrough came on 12 June 2024 with Dream Machine, a text-to-video model that turned prompts and still images into short clips and became one of the most widely shared video-generation tools of the year, alongside OpenAI's Sora. A developer API followed in September 2024, and in December 2024 the company raised a US$90 million round led by Amazon and AMD.^[3][7]

The Ray series arrived in 2025: Ray2 in January, a partnership with Saudi technology company HUMAIN in May, and Ray3 on 18 September 2025, which Luma billed as the world's first reasoning video model and the first to generate true high-dynamic-range video for professional use. In November 2025 Luma raised a US$900 million Series C led by HUMAIN — a company owned by Saudi Arabia's Public Investment Fund — with participation from AMD's venture arm, valuing Luma above US$4 billion, alongside the announcement of Project Halo, a two-gigawatt AI supercluster in Saudi Arabia.^[2][4][7]

Luma kept shipping through 2026: Ray3.14 in January brought native 1080p output at lower cost and higher speed, Ray3.2 in June added finer creative control, and the company refocused its products around Luma Agents — creative agents that coordinate multiple image and video models, including third-party systems such as ByteDance's Seedance 2.5, across a production workflow.^[5][6]

Technology

Luma describes its research goal as "unified intelligence" — multimodal foundation models that generate, understand and operate on visual content. The Ray models accept text, images and video as inputs and apply a reasoning step before generation; Ray3 outputs high-dynamic-range video suitable for professional post-production, and the company says it can be directed in ways that resemble directing a camera crew. Distribution runs through a web application, a developer API and integrations including Adobe Firefly.^[2][4][6]

Applications and impact

The technology is used in advertising, film and social media production. Agencies including Germany's Serviceplan Group adopted the platform in 2026, and a South African agency produced a Mazda commercial using Luma Agents. Luma says Dream Machine has served more than 30 million users, and its US$900 million raise illustrates both investor appetite for video generation and the rise of sovereign AI programmes in which states fund domestic compute and model capacity — the round was led by a Saudi state-owned investor. Luma competes with OpenAI's Sora, Google's Veo, Kuaishou's Kling AI and ByteDance's Seedance in an increasingly crowded video-generation market.^[1][2][7]

>See Also

• AI Video Generation
• Sora
• Veo
• Seedance
• Runway ML
• Adobe Firefly