Visual reasoning 2

Luma's first Video AI for reasoning, creating high-end HDR videos for film and advertising. Ray 3 is Luma's first Video AI for Reasoning, designed to produce videos in true EXR 10, 12, 12, 12, and 16-bit HDR formats, making it suitable for high-end film and advertising projects. It offers intelligent creation processes, including visual reasoning and HDR generation. Ray3 understands creative intent through advanced reasoning, interprets prompts with nuance using Chain of Thought, and allows users to export videos in 16-bit HDR or as EXR frames for professional workflows. It features state-of-the-art visual intelligence, including 16-bit HDR generation, Draft Mode for faster iteration, visual annotation, and advanced physics simulations for superior realism and character consistency.

Unified AI model combining visual reasoning and image generation in a single transformer. UNI-1 is a groundbreaking unified intelligence model that combines visual reasoning and image generation within a single decoder-only transformer architecture. Unlike traditional AI pipelines that separate understanding and creation, UNI-1 handles both in one model, eliminating modality gaps and enabling coherent multi-turn creative workflows. It is designed to perform structured internal reasoning—planning composition and resolving constraints—before rendering pixels. UNI-1 outperforms major competitors like Google and OpenAI on reasoning benchmarks such as RISEBench while offering significantly lower costs for high-resolution outputs.