EXCLUSIVE: Luma launches creative AI agents powered by its new ‘Unified Intelligence’ models
Luma's new agents can turn a $15M ad campaign into localized ads in 40 hours for under $20k.
AI video-generation startup Luma has launched Luma Agents, a new platform powered by its proprietary 'Unified Intelligence' family of models, designed to handle complete creative workflows across text, image, video, and audio. The system is built on the Uni-1 model, a multimodal architecture trained on a single reasoning system for audio, video, image, language, and spatial data. CEO Amit Jain describes its capability as 'intelligence in pixels,' where the model can understand and render creative concepts end-to-end. The agents are already being used by major brands like Adidas, Mazda, and global ad agencies Publicis Groupe and Serviceplan, moving beyond being a mere tool to fundamentally 'redoing how business is done.'
The key differentiator is the agent's ability to maintain persistent context and perform iterative self-critique, evaluating and refining its own outputs—a capability that has proven transformative in coding agents. Unlike current fragmented workflows that require mastering prompts for dozens of separate models, Luma Agents generate large variations and let users steer direction through conversation. In a compelling demonstration, the system turned a brand's year-long, $15 million ad campaign into multiple localized versions for different countries in just 40 hours for under $20,000, passing the brand's internal quality checks. While now available via API, Luma plans a gradual rollout to ensure reliability, positioning its agents as a game-changer for accelerating creative production in marketing and design.
- Powered by the new Uni-1 model, the first in Luma's 'Unified Intelligence' family trained on multimodal data for reasoning and generation.
- Agents coordinate with external AI models including Google's Veo 3, ByteDance's Seedream, and ElevenLabs' voice models for comprehensive asset creation.
- Demonstrated ability to localize a $15M ad campaign in 40 hours for under $20k, with self-critique loops and persistent context across iterations.
Why It Matters
This could drastically reduce the time and cost of large-scale creative campaigns for agencies and brands, moving from tool usage to workflow transformation.