--- title: "Ray3: Reasoning-Driven AI Video Model" description: "Create, animate & innovate with Luma’s AI. Generate cinematic videos using text, images, or clips powered by Ray 3 and Ray 2 with Dream Machine technology." canonical: "https://lumalabs.ai/ray3" source: "https://lumalabs.ai/ray3.md" --- # Ray3: Reasoning-Driven AI Video Model Ray3 is our next-generation video model, designed in collaboration with creatives from the entertainment, advertising, and gaming industries. With **reasoning-driven generation**, **best-in-class video-to-video** including character reference and keyframes, **Draft Mode** for rapid exploration and a **first-to-market HDR pipeline**, Ray3 brings consistency, creative control and production-grade fidelity to AI video. _NEW_ The latest update, **Ray3.14**, adds native 1080p generation, **4x faster performance**, improved stability, stronger prompt adherence, and **3x lower cost. [Learn more](https://lumalabs.ai/blog/news/ray3_14)** [Try for Free in Dream Machine](https://dream-machine.lumalabs.ai) ## Our Most Powerful Model - Fidelity - Physics - Adherence ### HDR Strikingly beautiful native 16bit High Dynamic Range generations. ### Reasoning Ray3 can think in concepts and visuals, understand intent, evaluate itself, and iterate to deliver better generations in fewer tries. ### Annotation Draw on images to get precisely what you want in the right place, with the right motion, and iterate without prompt engineering. ### Draft Mode An all new way to discover the perfect idea, in near realtime. 5x faster, 5x cheaper. ### SOTA Physics & Controls Ray3 brings incredible realism, accurate physics, improves image-to-video, keyframes, and controls. ### HiFi A new technique to master draft videos into production-grade 4K HDR. ### New — Modify Video Modify the first and last frame for cleaner, more controlled transitions. ### New — Character Reference Create consistent characters across shots with a single reference image. ## New — Modify Video Modify Video is drastically improved and now better maintains the physical logic, narrative coherence, and performance authenticity of your original footage. It supports layered, scene-aware transformations that behave as though they were captured naturally in camera. [Learn more](https://lumalabs.ai/blog/news/ray3-modify) - Wardrobe Swap - Environment Swap - Relighting - Virtual Product Placement ## New — Modify with Keyframes Now introduces Start and End Frame control to the video-to-video workflow for the first time. This allows you to guide transitions, control character behavior, and maintain spatial continuity across longer camera movement passes, reveals, and complex scene blocking. [Learn more](https://lumalabs.ai/blog/news/ray3-modify) - Magic Transitions - Editing Reality - Virtual Crowd - Doorway Threshold ## New — Character Reference Apply any custom character identity onto your original performance: a crucial functionality for actor-led projects with AI. For the first time ever, the feature provides you the ability to lock likeness, costume, and identity continuity of a specific character across an entire modified clip. [Learn more](https://lumalabs.ai/blog/news/ray3-modify) ### Use Cases - Mythical Creatures - Human Characters ## Reasoning Creative work is deeply intellectual. AI for creatives should be intelligent. Ray3 is capable of **understanding** your intent, **thinking** by generating visuals and concepts, and **judging** its outputs to give you better results. Reasoning enables creating more complex scenes, intricate multi-step motion, and do all of it faster. ### Chain of Thought Interpret prompts with nuance, judges early drafts, and retries until your quality bar is met. ### Visual Annotation Ray3 can interpret visual annotations like a creative partner and enables you to precisely specify layout, motion and character interactions. ## HDR Ray3 is the world's first model to deliver studio-grade HDR through native high dynamic range color generation. Vivid and striking HDR videos can be created from text prompts, SDR images, as well as SDR video and exported as 16-bit EXR for seamless integration into pro workflows. ### HDR Generation Tone map HDR videos to the perfect look. ### EXR Export Export videos as 16-bit EXR frames for professional editing, color grading, and compositing pipelines. ### SDR to Generative HDR Upgrade generated or recorded standard videos into vivid and accurate HDR footage. - "Man with a gold fish in a fishbowl for a head riding a shark like a motorcycle through a neon lit neo Tokyo alley at night." [Download HDR](https://static.cdn-luma.com/files/51d0a583d7cf26d4/(5)%20HDR%20example_1.mp4) · [Download EXR](https://static.cdn-luma.com/files/b77022ef886f2165/HDR%20Example_1.zip) - "Astronaut with flash light drifting through a destroyed dark space station zero gravity." [Download HDR](https://static.cdn-luma.com/files/51d0a583d7cf26d4/(6)%20HDR%20example_2.mp4) · [Download EXR](https://static.cdn-luma.com/files/b77022ef886f2165/HDR%20Example_2.zip) - "Skeleton zombie pirates walking through the thing jungle at night holding oil lanterns." [Download HDR](https://static.cdn-luma.com/files/51d0a583d7cf26d4/(7)%20HDR%20example_3.mp4) · [Download EXR](https://static.cdn-luma.com/files/b77022ef886f2165/HDR%20Example_3.zip) - "Man welding the cockpit of a spaceship open in the desert at night." [Download HDR](https://static.cdn-luma.com/files/51d0a583d7cf26d4/(8)%20HDR%20example_4.mp4) · [Download EXR](https://static.cdn-luma.com/files/b77022ef886f2165/HDR%20Example_4.zip) ## SOTA Fidelity Ray3 brings state of the art realism, physics, character consistency, and far superior instruction following. Ray3 also packs significantly more detail in the same resolution compared with Ray2 producing crisp high-fidelity outputs. - High Octane Motion - Preserved Anatomy - Adherent Instruction Following - Advanced Physics & Simulations - World Exploration & Understanding - Complex Crowds & Scenarios - High Fidelity Nuance & Details - Light Interactivity & Reflections - Motion Blur & Natural Photorealism - Lens & Focal Optics - Character Expression - Spatial & Character Consistency ## Draft Mode Introducing a new way to iterate. _Fast_. Draft Mode enables you to explore ideas in a **state of flow** and get to that perfect shot for your story. Then, with Ray3's new Hi-Fi Diffusion "master" your best shots into production-ready high-fidelity 4K HDR footage. Video creation is fun again! - **5x Faster**: See ideas come to life in seconds. It’s like sketching videos. - **5x Cheaper**: Use your credits 5x more efficiently in Draft Mode and then Hi-Fi only the very best shots. ## Designed to Tell Stories ### Wasted After a motorcyclist scratches the wrong car he is thrust into a white-knuckle, high-octane chase as an unhinged driver turns petty revenge into a city-wide war on wheels. _Directed by Hahn · Presented by Dream Lab LA · Powered by Luma AI Ray3_ ### Subway The story begins in a subway and unfolds into a fantastic, musical journey where giant creatures, playful ideas, and impossible wonders collide. To capture this variety, we invited ten artists to contribute their wildest concepts, then shaped a unifying look that allowed all those visions to coexist. Edited together, the result is a cinematic piece as expansive, unpredictable, and alive as Ray3 itself. _Directed by Felipe Machado · Presented by Silverside · Powered by Luma AI Ray3_ ### Homesick In an apocalyptic near-future California, a lone vagabond must traverse threats across a dangerous urban hellscape once known as Los Angeles. _Directed by Fishkin · Presented by Dream Lab LA · Powered by Luma AI Ray3_ ### The Double Double The Double Double is a wild, absurdist comedy where diner booths double as cosmic battlegrounds. Four mismatched friends—stand-ins for an unstable star system—can’t drift farther than sixty “waffle units” apart without triggering galactic collapse. When egos clash and someone crosses the line, syrup-drenched meteors, cosmic dogs, and the laws of gravity itself unravel in spectacular chaos. _Directed by Verena Puhm · Presented by Dream Lab LA · Powered by Luma AI Ray3_ ## Feature Gallery ### Text-to-Video - "Two people in the eye of a tornado, one is being sucked up in the tornado, the other one tries to catch them but their hands let loose, cinematic." - "A girl rides a massive, majestic fish through the ocean's depths. Sunlight illuminates shimmering scales as they glide through the deep blue. A magical scene of freedom and wonder." - "A woman rides a bear running rapidly through a forest in a mountainous area, wide shot, motion blur." - "A super slow mo macro anamorphic bokeh of a water bender silhouetted child with beautiful eyes illuminated with highlights and blue floating water in the background and foreground backlit in slow motion." - "A massive iceberg dramatically collapses into the ocean. Jagged, blue shards tumble in slow motion, scattering frozen debris. Shimmering light reflects off sharp edges amidst mist and icy spray. " - "Golden hour illuminates a small bird perched on a resting rhino's horn in a hyper-realistic, cinematic close-up. The camera intimately captures the rhino's weathered skin and the bird's delicate movements in a symbiotic act. A blurred savanna and soft breeze complete the scene." - "In a quiet, natural setting, a translucent celestial being interacts gently with a solitary human. The being shimmers with starlight, galaxies swirling within its semi-transparent form. The human, in awe, stands still as the being extends light, creating ripples of starlight. A cosmic hum underscores this intimate, reverent connection between the infinite and the mortal." - "A close-up video of a sorrowful woman in a mystical forest. She slowly lifts her head upward, tears streaming down her cheeks. The tears glisten with green glitter shimmer, sparkling faintly as they catch the light. She wears an ancient helmet made of bark, moss, and vines, blending her with the forest itself." ### Image-to-Video - - - - - - - - ## Available now in Dream Machine. [Try Now](https://dream-machine.lumalabs.ai)