Luma and Dumbstruck launch Creative Intelligence, pairing emotion analytics with AI video editing

Luma and emotion-analytics company Dumbstruck have launched a joint product called Creative Intelligence, combining audience-response measurement with Luma's AI video editing agents to let advertisers test and refine video ads before they ever run, the companies announced on August 13, 2026.

What's new

• Dumbstruck's platform measures how audiences respond to advertising content — emotionally, behaviorally, and cognitively — and converts those responses into specific feedback about what's working, what isn't, and where content can improve.
• Luma Agents then execute the identified changes directly on the video, using frame-by-frame editing control, without requiring a reshoot or a manual edit pass.
• The two capabilities form a closed loop: content is measured, refined by Luma's agents based on that measurement, and can be re-tested before it goes to market.
• Dumbstruck CEO Jeff Tetrault framed the underlying problem the product targets: "AI is changing the economics of advertising creative. The challenge isn't creating more. It's deciding what deserves to exist."
• Luma COO Caroline Ingeborn described the intended effect: "Human creativity is not the bottleneck here — uncertainty is. Creative Intelligence removes that uncertainty."

Context

The launch responds to a problem created by generative AI itself: as AI video tools make producing large volumes of ad creative cheap and fast, the bottleneck shifts from production capacity to knowing which of many possible versions will actually perform with a target audience. Dumbstruck's core business — emotion and attention analytics for advertising — predates the current wave of generative video tools, but pairing it directly with an AI agent that can act on its findings is new. Rather than a marketer manually interpreting analytics and re-briefing a creative team or a separate AI tool, Creative Intelligence closes the loop programmatically: measurement and revision happen inside one connected workflow.

This also fits a broader pattern in AI video tools this year, where vendors like Luma have increasingly positioned generation as one part of a larger production and optimization pipeline, rather than a standalone output. Competing video generation platforms have leaned mostly into raw generation quality and speed; this partnership instead targets the layer above generation — deciding which generated content is worth shipping.

Why it matters

For advertisers and agencies already producing high volumes of AI-generated video creative, the practical bottleneck Luma and Dumbstruck are targeting is real: more generated variants only help if there's a fast, reliable way to know which ones will land with an audience before spending media dollars against them. A tool that both predicts performance and can act on that prediction — without a manual reshoot — shortens the test-and-iterate cycle from days to something closer to real time. It's a narrow, ad-tech-specific launch rather than a frontier model release, but it's a concrete example of AI video tooling moving up the value chain from "generate the clip" to "decide whether the clip should exist," which is where much of the commercial value in AI-generated advertising is likely to concentrate as raw generation quality becomes commoditized across vendors.