Luma’s New AI Agents Handle Text, Image, Video, and Audio Creation

AI startup Luma on Thursday unveiled Luma Agents, a platform that can plan and generate creative content across text, images, video, and audio. Powered by the company’s Unified Intelligence models, the system is designed for agencies, marketing teams, and design studios that need to produce and manage multimodal creative work inside a single platform.

The new agents run on Luma’s Uni-1 model, the first model in the company’s Unified Intelligence architecture. According to Luma CEO and co-founder Amit Jain, Uni-1 was trained on audio, video, images, language, and spatial reasoning so it can plan ideas in text and then generate visual outputs from the same underlying reasoning system. Jain describes this approach as giving the model the ability to “think in language” while rendering results visually.

Luma Agents can also coordinate with other AI models rather than relying only on Luma’s own tools. The platform can integrate with systems such as Luma’s Ray 3.14, Google’s Veo 3 and Nano Banana Pro, ByteDance’s Seedream, and ElevenLabs voice models, allowing the agents to generate or refine content across different formats while managing the workflow between them.

The system is aimed at reducing the back-and-forth prompting that often defines creative AI workflows today. Instead of requiring users to manually iterate on each step of a project, the agents can generate multiple variations of ideas and refine them through an internal feedback loop that evaluates and improves the outputs. Jain said this self-evaluation capability—similar to what has made coding agents effective—allows the system to iterate until results meet the desired standard.

Luma has already begun deploying the platform with existing customers, including Publicis Groupe and Serviceplan, as well as brands such as Adidas, Mazda, and the Saudi AI company Humain. Jain said the agents can maintain persistent context across creative assets, collaborators, and project iterations, allowing teams to manage complex campaigns within a single workflow.

In a demonstration, Luma showed how a brief product description and an image could produce multiple advertising concepts, including variations in settings, models, and color palettes. Jain also said the system was able to convert a brand’s $15 million, year-long advertising campaign into localized ads for multiple countries in about 40 hours at a cost of under $20,000, while still meeting the brand’s internal quality checks.

Luma Agents are now available through an API, though Jain said access will roll out gradually to ensure reliability as more teams begin integrating the system into their creative workflows.

See more here:

This analysis is based on reporting from TechCrunch.

Image courtesy of Luma.

This article was generated with AI assistance and reviewed for accuracy and quality.