Luma AI

Luma AI builds models and software for generating and editing visual media. Its Ray video models turn text and images into clips, transform existing footage, and support keyframe direction and HDR output. UNI handles image generation and editing with reference images and natural-language instructions. In Luma App, creative agents plan, generate and refine work across video, images and audio, sharing project context across teams. Developers can integrate the models through APIs, while advertising, film and gaming teams use the tools for campaigns, sequences and production workflows.

Founded in 2021, Luma counts Amit Jain and Alex Yu among its co-founders; Jain serves as CEO. Its early generative 3D work included Genie, which made interactive 3D assets from text prompts. Its research also addresses the computational cost of generation: Terminal Velocity Matching generalizes flow matching to generate outputs in one or a few steps, combining architectural changes and a specialized attention kernel to support stable training.

Luma sells individual workspace subscriptions, usage-based API access, dedicated capacity and enterprise services. Its historical Dream Machine platform attracted one million users within four days of launching in 2024. In 2025, Luma raised a $900 million Series C led by HUMAIN to expand research and engineering. Luma also partnered with HUMAIN on Project Halo, a planned compute cluster for training and inference.