What Luma AI does
Luma AI is a multimodal AI company building “world models” and generative video/3D systems for creating, understanding, and operating in the physical world. Its consumer-facing flagship platform, Dream Machine, provides interactive video generation and creative tools, while its research and model line is centered around reasoning-oriented video models (Ray series) and fast generative 3D creation (Genie). In parallel, Luma is also positioning for large-scale multimodal model training and deployment—most recently via Project Halo, a compute supercluster partnership announced alongside its $900M Series C.
In practice, Luma’s products aim to move beyond prompt-to-output into controllable creative workflows (e.g., “Modify”/frame-level direction and multi-step “reasoning” video generation), and to make 3D assets/worldbuilding accessible at scale (Genie and related capture/edit experiences). Luma’s business model is primarily a paid subscription/API approach (the Dream Machine platform is available for creators and enterprises, and Luma also provides access via API/enterprise offerings). Luma’s most visible differentiation is its focus on physically grounded multimodal intelligence for creative production—combining video generation fidelity (including HDR EXR output described for Ray3) with editing/control features that are designed for downstream creative pipelines.
Strategically, Luma has moved from early 3D capture and NeRF-style workflows into frontier multimodal generation and reasoning video, while scaling compute through partnerships (AMD, HUMAIN) and expanding commercial distribution via major creative/enterprise partners (including Adobe for Firefly integration and agency/production ecosystem partnerships).