Luma launches AI assistants with new ‘Unified Intelligence’ models

AI video startup Luma on Thursday launched Luma Agents, designed to handle end-to-end production tasks for audio, images, videos, and audio. Luma Agents are supported by the founders of the Unified Intelligence family of models, with built-in architectures trained on a single multi-intelligence system.

Luma Agents is positioned as a new way of working for advertising agencies, marketing teams, production studios, and businesses. Luma says its assistants can process and create text, images, videos, and audio while interacting with other AI models, including Luma’s Ray 3.14, Google’s Veo 3 and Nano Banana Pro, ByteDance’s Seedream, and ElevenLabs’ voice models.

Luma’s agents are built on the original Uni-1 model, the first of its Unified Intelligence family of AI models. They were trained in audio, video, graphics, language, and spatial reasoning, according to Amit Jain, CEO and co-founder of Luma.

Jain told TechCrunch that the Uni-1 model can “think in language and imagine and interpret pixels or images … we call it ‘intelligence in pixels.’

“Our customers are not buying a tool; they are redefining the way business is done,” Jain said.

Luma has already started rolling out its new platform with existing clients, including global advertising agencies Publicis Groupe and Serviceplan, as well as brands like Adidas, Mazda, and Saudi AI company Humain.

Jain said Luma Agents are a game changer because they can keep things stable in terms of finance, support, and technology iteration. They can also evaluate and revise results, improving their results through repeated self-evaluation, according to Jain.

That kind of ability is what makes coding so useful, Jain said. “You need this skill to test your work, improve it, and take action until the answer is good and correct.”

Jain said the current trend of using AI tools in the manufacturing environment does not have the same acceleration of benefits that people in the manufacturing industry expect from AI. Instead, it’s like: “Here are 100 examples. Learn how to help them,” he said.

He said that what makes Luma Agents different is that you don’t have to run back and forth for each iteration of an image or idea – the system instead creates different large models and allows users to control the process through dialogue.

“With Unified Intelligence, because these models understand in addition to being able to create, we can create a system that can do this kind of work,” Jain said.

For example, consider a person who draws house plans. As he draws the lines, he is creating an internal mental image of the structure, light, energy of the place, and life events. This, Jain says, is the very principle on which Unified Intelligence is built.

Jain said the system could significantly speed up manufacturing operations. In the presentation, they showed how 200 short words and an image of a product (a tube of lipstick) led the system to generate different ideas for locations, models, and brand plans for an advertising campaign.

In another example, Luma Agents turned a $15 million, year-long brand advertising campaign into multiple international ads in 40 hours for less than $20,000, bypassing the brand’s internal controls and checking for accuracy, Jain said.

Although Luma Agents are now publicly available through an API, Jain said the startup plans to roll out access gradually to ensure users have reliable access and avoid workflow disruptions.