A Tencent veteran is taking on OpenAI's Sora with a decidedly different pitch: forget viral memes, think movie studios.
Two months after incorporation, Video Rebirth has already pulled in $50 million—the kind of seed round that makes seasoned venture watchers do a double-take.
The Singapore-based startup announced the funding November 5, though Bloomberg had to do the work of identifying the backers: Qiming Venture Partners and South Korean gaming outfit Actoz Soft. Video Rebirth itself kept things vague in its official statement, naming only "top-tier global financial and strategic investors." That reticence may signal ongoing negotiations, or simply a preference for stealth mode while the product remains under wraps.
What's less mysterious is the pedigree. Dr. Wei Liu, the company's founder, spent years as a Distinguished Scientist at Tencent, where he helped architect the tech giant's Hunyuan foundation model. An IEEE and AAAS Fellow, Liu knows his way around large-scale AI infrastructure—credentials that likely made the pitch deck easier to sell, even with barely eight weeks on the corporate registry.
The Physics Problem
Video Rebirth's central argument is straightforward enough: current AI video generators are parlor tricks masquerading as production tools. Sure, they'll give you something eye-catching for social media. But try using them for a commercial shoot or an animated film, and the cracks show fast.
Physics breaks. Shadows float disconnected from their objects. Motion becomes soup.
Enter what the company calls "Physics Native Attention," the proprietary architecture underlying its "Bach" series of models. (The name presumably nods to the composer, though Video Rebirth hasn't explained the reference.) The technology purports to solve the consistency problems that plague competitors—realistic movement, coherent lighting, the subtle interplay between objects that signals "this wasn't made by a hallucinating algorithm."
Liu frames the mission in world-model terms, describing the goal as a "video-native world model" capable of cinematic output without requiring armies of post-production artists to clean up the mess. Whether Physics Native Attention actually delivers on that promise remains an open question. The company hasn't released demos, benchmarks, or side-by-side comparisons. For now, it's a technical claim awaiting validation.
Betting Against the Consumer Play
Here's where Video Rebirth diverges sharply from most of the AI video pack: it's explicitly not chasing TikTok creators or weekend hobbyists.
"We're focused on professional creators who need high-fidelity, controllable output," the company stated in its funding announcement. Translation: advertising agencies, e-commerce studios, film production houses, animation teams. The workflow integrations matter more than the wow factor.
That positioning puts Video Rebirth in a curious spot. The consumer AI video market is noisy and well-capitalized—Runway pulled in $308 million this past April at a valuation north of $3 billion, with General Atlantic, SoftBank, and Nvidia all writing checks. Pika Labs has raised roughly $135 million. These companies have marketing momentum and user bases to show for it.
Video Rebirth, by contrast, is making a contrarian wager: that there's a lucrative gap between novelty tools and production-grade technology, and that professional buyers will pay a premium for reliability over virality. It's not an absurd thesis. Hollywood studios and Fortune 500 marketing departments have deeper pockets than influencers, and their tolerance for glitchy output is considerably lower.
But it's also a tougher sale. Enterprise customers move slowly, demand extensive proof-of-concept work, and expect white-glove service. Video Rebirth will need to show that its models don't just sound impressive in a pitch deck—they need to survive contact with real production deadlines.
What Comes Next
The company plans to ship its 1.0 product in December, which leaves roughly a month (perhaps less, depending on when you're reading this) for the team to make good on the technical promises. The $50 million will fund both R&D sprints and the buildout of what Video Rebirth describes as an "AI Generated Entertainment" ecosystem—a term vague enough to mean almost anything, from plugin marketplaces to partnership deals with post-production software vendors.
For now, Video Rebirth's web presence is almost comically sparse: a single tagline ("From video generation to video-native World Model"), a contact form, and not much else. No product screenshots. No waitlist hype. No founder blog posts explaining the vision in accessible terms.
That minimalism could be strategic—why show your hand before the product is ready?—or it could reflect a team still figuring out how to talk to customers outside the AI research community. Either way, December will clarify whether this $50 million bet was prescient or premature.
In a market where every other startup claims to be "revolutionizing video," Video Rebirth is at least asking a different question: what if the revolution isn't about making more AI video, but making AI video that professionals can actually use?
The answer, as with most things in early-stage tech, will come down to whether the models work as advertised. Pedigree opens doors. Product performance keeps them open.
