Text-to-video today: type a line, wait ten minutes, get ten seconds of footage. Reactor aims to kill that wait.
On May 28, the company emerged from stealth, announcing a $59 million Series A round led by Lightspeed, with participation from Katzenberg's WndrCo — yes, the DreamWorks Katzenberg.
What Makes It Different
The key word is "real-time." Ordinary text-to-video is batch processing: you submit a request, it renders in the background, then delivers. Reactor takes a different path — frames are generated on the fly, with no length limit, and you can interact continuously.
Founder Alberto Taiuti said: "With our platform, the latency to the first frame is essentially zero."
This is powered by a "world model." Unlike models that just spit out a static video, a world model perceives and reacts in real time — if you move left, it builds the left-side scene on the spot.
Sound familiar? Last week Google's Genie did something similar: you click a point on a map, and it builds a world for you to walk into. The difference is Reactor doesn't make its own apps; it packages the capability into an SDK and API that can be integrated with just a few lines of code, for developers to use.
Who's Behind It
CEO Alberto Taiuti and CTO Bryce Schmidtchen were both technical leads on Apple's Vision Pro. Taiuti also has a more notable background: he is a co-founder and former CTO of Luma AI, a company whose 3D and video generation tools are widely used. In other words, he has built real-time generation systems before.
The company was founded in August last year, based in San Francisco's South Park, with 16 people — all with impressive backgrounds from Apple, Netflix, Meta, Google, Adobe, Replicate, and Microsoft.
For compute, AWS is its preferred cloud. Already using the platform is an AI lab called Overworld, along with a dozen model teams.
Why Investors Are Betting Now
Because the direction has shifted. Over the past two years, the competition was about "who can generate clearer, longer, more realistic video" — a static content comparison. Reactor is betting on the next phase: video that is no longer just for watching, but for walking into and interacting with.
Gaming, robotics, and so-called "physical AI" all need a real-time responsive visual engine. Whoever crosses the "real-time" threshold first will lock in the gateway to these scenarios.
Schmidtchen describes the company as "a bridge between the world of models and real-world applications." Katzenberg's verdict after investing: he was "genuinely shaken" by the novelty of the approach.
Still a Gap
But there's cold water to pour: Reactor hasn't disclosed pricing details, and how it actually performs can only be heard from its own claims for now.
The real-time world model direction sounds sexy, but "real-time" and "quality" often conflict — fast tends to be rough, refined tends to be slow. Reactor says it wants both, but that claim will only count when real products are on the table.
It has opened its SDK and API today. Developers will soon vote with their feet.
Sources: Real-Time AI Video Startup Reactor, Founded by Ex-Apple Engineers, Raises $59 Million (Variety); CocoLoop; Reactor Emerges from Stealth with $59M to Build the Platform for Real-Time AI Worlds (PR Newswire)