About Decart
Decart was founded by Dean Leitersdorf and a team of AI researchers and infrastructure engineers, pursuing a technically distinctive niche within the broader generative AI video race: rather than building models that generate short, pre-rendered video clips from a text prompt, a process that can take anywhere from seconds to minutes per clip, Decart has focused specifically on real-time, continuous video generation and editing that can run live, indefinitely, and respond instantly to changing input, whether that's a live camera feed, a viewer's prompt, or a player's keyboard input in an interactive environment.
The company's technology stack, which it calls DOS, is a heavily optimized inference and training system built to minimize latency and cost for this kind of always-on, real-time generative video workload, and its two primary model families are Lucy, aimed at immersive real-time video experiences such as live-streamed AI-generated visual effects and character transformations, and Oasis, aimed at interactive, playable "world models" that generate video game-like environments frame by frame based on user input, an early example of AI potentially generating playable game worlds on the fly rather than relying on pre-built game engines and assets. Decart has demonstrated its technology with live Twitch streamers, performing real-time visual transformations during live broadcasts, and has partnered with Amazon Web Services to run its models on AWS's custom Trainium AI chips, part of a broader industry trend of AI companies diversifying away from exclusive reliance on Nvidia GPUs. The company has raised substantial and rapidly escalating venture funding, including a 2025 round reportedly valuing it at roughly $3.1 billion and a subsequent 2026 round led by Radical Ventures at a valuation approaching $4 billion, with participation from Nvidia, Adobe Ventures, Sequoia Capital, and Benchmark, reflecting significant investor interest in real-time generative video and world model technology as a distinct frontier from mainstream text-to-video AI.
