Luma Dream Machine is a world-model AI video platform powered by Ray 3, generating physically accurate video with character consistency and cinematic quality.
Luma AI builds multimodal general intelligence, and its Dream Machine platform puts that research directly in creators’ hands. The Ray 3 video model, the world’s first reasoning video model, simulates physics, motion, and object permanence rather than just hallucinating pixels. Water splashes and car drifts follow calculated trajectories. Character likeness, costume, and identity lock across entire shots, eliminating the morphing glitches of earlier AI video.
Ray3 Modify introduced hybrid-AI workflows: film a real actor’s performance, then guide scene evolution with AI, preserving motion, timing, eye line, and emotional delivery while transforming locations, costumes, and styles. Backed by a $900 million round led by Humain with a16z and Amazon, Luma partners with Adobe and AWS, serving entertainment studios and advertising agencies. The platform also includes Photon, a high-fidelity image model, and Uni-1 for brand-level visual intelligence.
Free plan includes watermarked personal-use generations. Lite and Standard plans are credit-based, with 5-second generations costing 170 credits. Plus and Unlimited remove watermarks and enable commercial use. Unlimited plans support heavy production use. API pricing is available for developers.
Filmmakers, agencies, and studios needing realistic AI video with character consistency. Content creators ideating mood films and social clips. Casual users may find the credit math complex; Pika suits quick experiments better.