Dean Leitersdorf is the latest to jump into the AI circus, showcasing a model from his startup, Decart, that’s meant to manipulate live video in real-time. He types in prompts like “wild west, cosmic, Roman Empire” and, unsurprisingly, it churns out a surreal version of Julius Caesar underwater. Sure, it’s a neat trick, but anyone who’s been around the block knows that tech demos are often a far cry from practical, scalable products.
Leitersdorf, channeling the archetypal tech visionary with his long hair and eccentric demeanor, demos a model that morphs his appearance with each prompt. It’s entertaining, but let’s not kid ourselves—this isn’t the first time we’ve seen this kind of tech spectacle. The real question is whether it has staying power or is just another flash in the pan.
Decart’s model, Mirage, claims to be an engineering marvel, touted as a disruptor for the livestreaming industry. On paper, it sounds like a game-changer, but remember, we’ve heard similar promises before. The tech can manipulate video with simple text prompts, but whether it can handle more complex, real-world applications remains to be seen.
This Thursday, Decart is launching a platform for users to create and modify videos. They’re offering default themes like “anime” and “cyberpunk,” and while it sounds flashy, let’s not confuse novelty with innovation. Transforming a Fortnite scene into an underwater world is amusing, but we should be asking if it adds meaningful value or is just a gimmick.
Decart’s technology might make waves in gaming. They demoed a game called Oasis that generated a Minecraft-like world on the fly, which sounds impressive if it can be executed consistently. However, we’ve seen hyped gaming tech before that never quite delivered on its lofty promises.
The real-time manipulation of video is computationally demanding. Decart has optimized their model to run on Nvidia chips, achieving a modest 20 frames per second. This might be enough for TikTok clips, but it’s hardly groundbreaking. Until they can deliver full HD or 4K output reliably, consider me skeptical.
Building coherent video in real-time is a slippery slope, with potential for the model to veer off course. Decart has developed methods to maintain coherence, but let’s be clear: correcting errors at speed is easier said than done. They’re working on more advanced features, but until they can demonstrate solid results, it’s just talk.
The tool might gain traction on social platforms like TikTok or Instagram due to its novelty. However, its unpredictability—like randomly altering a user’s race—could be a sticking point. These kinds of issues are less amusing and more problematic in the real world.
Leitersdorf claims only major AI players have the technical chops to build something like Mirage, and insists he’s not looking to be acquired. Instead, he’s aiming for a “kilo-unicorn”—valued at a trillion dollars or a billion users. Ambitious? Sure. Realistic? That’s another story. Given the volatile nature of tech startups, it’s best to keep expectations grounded until they prove they can deliver on these grand visions.
