Luma AI launches Ray3, the video generation model for next-generation movies with inference built-in

Artificial intelligence startup Luma AI Inc. today announced the launch of Ray3, a powerful text-to-video AI model with built-in reasoning designed for high-quality cinema visual production for professionals.

Luma has also revealed its partnership with Adobe Inc. to integrate the new model into Adobe's AI-enabled Firefly app, all-in-one software for creative work.

“Ray3 is the first step in building intelligence for creative work,” said Amit Jain, co-founder and CEO of Luma AI. “Creative work is one of the most intellectually challenging things humans do, but up until now, many of the AI ​​that creatives can use have fallen behind what is possible in coding and analytics in language models.”

The flagship feature of Ray3 is reasoning in mindsets so that you can “think” through scene descriptions and follow instructions from creative experts. Jain said the most generated video models on the market so far are similar to slot machines.

Inference allows Ray3 to evaluate its own output and improve the results to better maintain the user's artistic vision. You can plan complex scenes and determine if the output makes sense before presenting them.

This model works in a similar way to animators and filmmakers, sketching the storyboard before generating the final product. During this drafting process, users can work with the model to provide more accurate instructions, such as annotating portions of the video. The model can then follow complex, multi-step ideas. Additionally, visual annotations are understood, such as lines drawn on video stills during draft, allowing users to follow more accurately the instructions.

https://www.youtube.com/watch?v=z-ahv1jpz6u

The Ray3 represents a major upgrade to the company's previous generation model, the Ray2, twice the size. You can generate true high dynamic range videos using the Professional ACES2065-1 EXR standard in 10, 12, and 16-bit formats. In fact, this gives filmmakers and advertisers access to the same range of color exposure and lighting controls in footage shot with high-end cameras.

This model can also employ standard dynamic range videos from virtually any source, converting them to HDR, providing richer color and editing flexibility. For example, Ray3's HDR conversion can brighten up excessively dark scenes without “washing away” the colors.

Users can generate video clips of up to 10 seconds from both text and images. By annotating text on the image, the user has even stronger control over the initial output. Thanks to the powerful configuration and visual understanding engine built into the model, sewing together multiple scenes is easier as it maintains generational consistency than before.

In addition to its partnership with Adobe, Luma Ai said that Ray3 is being adopted by Dentsu Digital Inc., one of Japan's largest integrated digital marketing companies. As a launch partner, Dentsu plans to use Ray3 in its production pipeline to provide domestic brands with greater control and ability to personalize and storytelling.

Creative leaders, including digital marketing company Monks and advertising company StrawberryFrog LLC, are also expanding their capabilities with Ray3. Additionally, Saudi AI company Humain said it plans to integrate Ray3 into an enterprise service aimed at creative professionals.

“Ray3 is more than just an upgrade, it's a quantum leap,” said Steve Plimsoll, Chief Strategy Officer at Humain. “By giving AI the power to reason beyond words, images and movement, we are not only woven beyond the speed and fidelity of creative output, but also intelligent guardrails. It means faster, more keen ideas and safer content that respects ethics, compliance and cultural context.”

Image: Luma AI