Luma AI has launched a "Unified Intelligence" architecture designed to solve the "memory problem" and fragmentation of current AI tools by creating a single multimodal system that can reason and generate content across text, images, and video without losing context.
Its new product, Luma Agents, are AI tools built on this architecture to streamline creative production for professionals, reportedly reducing a campaign that would take months and millions of dollars to just 40 hours of artist work.
The company, which recently raised $900 million, is targeting the global creative industry by offering full IP ownership and positioning its platform as operational infrastructure to help teams scale.
Main Topics: Luma AI's Unified Intelligence architecture and new AI agents, the "memory problem" in existing AI, the impact on creative workflows, and the company's recent funding.
Luma AI, a Palo Alto-based startup that turns prompts into videos, says it's time to fix the fragmented nature of today's AI tools which forget details and hinder creative work.
Luma's Unified Intelligence architecture, launched on Thursday, claims to fix this with one multimodal AI brain that blends reasoning with content generation and can handle text, images, video, and more in a single system, without losing track between steps. Its first model, Uni-1, mixes words and visuals together so the AI can think through ideas and create them at the same time.
The startup launched Ray3 in 2025 as the world's first video reasoning model, followed by Ray3.14, which creates videos, animations, and visuals.
Last November, Luma raised $900 million in a round led by Humain, a Saudi-based public investment fund (PIF), with participation from existing investors Andreessen Horowitz, Amplify Partners, and Matrix Partners.
Most AI models process only a fixed amount of prior conversation or data at a time, causing them to "forget" earlier details in long projects like coding marathons or storyboarding, requiring constant re-prompting. Big tech companies have repeatedly flagged AI's "memory problem" which leads to incoherent output, higher cost, and failed projects.
Luma also launched Luma Agents, a new class of AI tools built on its new architecture, to enable creative work for agencies, marketing teams, studios, and enterprises.
Jain said these agents have the full picture across text, images, video, and audio. They look for different creative ideas, check and improve their own work (like coding agents), and connect easily to other tools through APIs.
The agents work on external models such as Veo 3, Sora 2, Kling 2.6, Nano Banana Pro, and ElevenLabs and help cut down production timelines. Jain narrated that using Lumaâs agents, it took two artists 40 hours to work on a campaign that would otherwise have taken 12-13 months and $12 million.
Jain explained that this represents a fundamental shift for advertising, marketing, and media workflows since the agents enable end-to-end production. "Agents collaborate end to end... everything, from the cast, locations, the lighting, colours, shots, the scenes, the motion, and how the thing flows, rather than isolated tasks".
Creative work has never lacked ambition but execution capacity, with teams wasting time orchestrating tools instead of focussing on taste, direction, and strategy, he said. The startup will provide full IP ownership for customers, automated content review, and traceability, positioning its platform as an operational infrastructure that helps creative teams scale.
Luma is targeting a global workforce of 100 million creative professionals, including designers, architects, and filmmakers. With global ad spend poised to cross the $1 trillion mark this year, Jain feels the shift from isolated tools to unified agents is inevitable.
Luma's Unified Intelligence architecture, launched on Thursday, claims to fix this with one multimodal AI brain that blends reasoning with content generation and can handle text, images, video, and more in a single system, without losing track between steps. Its first model, Uni-1, mixes words and visuals together so the AI can think through ideas and create them at the same time.
The startup launched Ray3 in 2025 as the world's first video reasoning model, followed by Ray3.14, which creates videos, animations, and visuals.
Last November, Luma raised $900 million in a round led by Humain, a Saudi-based public investment fund (PIF), with participation from existing investors Andreessen Horowitz, Amplify Partners, and Matrix Partners.
Most AI models process only a fixed amount of prior conversation or data at a time, causing them to "forget" earlier details in long projects like coding marathons or storyboarding, requiring constant re-prompting. Big tech companies have repeatedly flagged AI's "memory problem" which leads to incoherent output, higher cost, and failed projects.
Luma also launched Luma Agents, a new class of AI tools built on its new architecture, to enable creative work for agencies, marketing teams, studios, and enterprises.
Jain said these agents have the full picture across text, images, video, and audio. They look for different creative ideas, check and improve their own work (like coding agents), and connect easily to other tools through APIs.
The agents work on external models such as Veo 3, Sora 2, Kling 2.6, Nano Banana Pro, and ElevenLabs and help cut down production timelines. Jain narrated that using Lumaâs agents, it took two artists 40 hours to work on a campaign that would otherwise have taken 12-13 months and $12 million.
Jain explained that this represents a fundamental shift for advertising, marketing, and media workflows since the agents enable end-to-end production. "Agents collaborate end to end... everything, from the cast, locations, the lighting, colours, shots, the scenes, the motion, and how the thing flows, rather than isolated tasks".
Creative work has never lacked ambition but execution capacity, with teams wasting time orchestrating tools instead of focussing on taste, direction, and strategy, he said. The startup will provide full IP ownership for customers, automated content review, and traceability, positioning its platform as an operational infrastructure that helps creative teams scale.
Luma is targeting a global workforce of 100 million creative professionals, including designers, architects, and filmmakers. With global ad spend poised to cross the $1 trillion mark this year, Jain feels the shift from isolated tools to unified agents is inevitable.
Meridiano
