Sequoia led a $10 million seed round into Preview six months after General Partnership’s $2 million pre-seed. Total capital: $12 million. Over 100 studios already onboard. 3,000 more queued. The numbers scream “hot deal.” But the real signal is buried in the architecture — not the dollar amount.
Preview isn’t just another AI video wrapper. It’s a central control panel for the entire production pipeline: scripts, storyboards, shot lists, AI generation, review, feedback — all in one workspace. Multiple models run simultaneously. Characters, scenes, props managed in a unified database. Each frame logs who generated it, which model was used, and the exact parameters applied.
Sequoia called it the “video version of Cursor.” Cursor is the code editor that replaced IDEs for AI-assisted development. Preview wants to be that for video production. But the comparison undersells the technical depth. Preview’s real innovation is the metadata layer — a chain of custody for every pixel.
Context: The Fragmentation Problem
AI video generation is currently a mess. A studio uses Midjourney for storyboards, Runway for keyframes, Pika for motion, ElevenLabs for voice, and DaVinci Resolve for final assembly. No unified standard. No shared asset library. Each tool spits out proprietary metadata — or none at all. Teams waste hours reconstructing what was used where.
I’ve seen this pattern before. In 2019, during the Telegram phishing campaign I reverse-engineered, the attackers had a similar problem: fragmented command-and-control infrastructure. They solved it by centralizing key logs. Preview does the same for production. It becomes the single source of truth for every creative decision.
Core: The Provenance Engine
Preview’s architecture is not just about convenience. It’s about forensic traceability. Every frame carries a digital birth certificate: generation timestamp, model ID, prompt, seed, negative prompt, resolution, steps, guidance scale. This is not a feature for most users. It’s a feature for the production pipeline that will eventually need to prove authenticity.
Think about it. Hollywood studios, Fortune 500 agencies — these clients face liability. If an AI-generated frame in a commercial accidentally includes copyrighted IP, the studio needs to prove it was generated from a specific prompt, not stolen. Preview’s metadata provides that proof. It’s a governance layer for creative assets.
From my experience analyzing the Yearn Finance governance proposal — where I traced tokenomics centralization risks — I learned that the absence of transparent records is the root of most disputes. Preview eliminates that ambiguity. It’s not just a tool; it’s a legal shield.
The Funding Velocity
$2 million pre-seed from General Partnership. Six months later, $10 million seed from Sequoia. That’s a 5x increase in valuation in half a year. Investors are betting on infrastructure, not just application. Sequoia’s track record with Cursor (which reached $100M+ ARR) suggests they see Preview as the next platform play.
But the waiting list is the real metric. 3,000 studios waiting. That’s not a sales funnel; it’s a demand signal. Studios are desperate for standardization. The fragmented AI video market is bleeding efficiency. Preview offers a unified command center.
Contrarian: The Centralization Trap
The obvious blind spot: Preview is a centralized platform. All metadata, all assets, all workflows live on their servers. That’s a single point of failure — both technical and political. If Preview goes down, thousands of productions halt. If they change pricing, studios have no leverage. If they get acquired, the data could be monetized against users.
In crypto, we call this “sequencer centralization.” Layer2 rollups face the same criticism: the sequencer controls the order of transactions. Preview is the sequencer for AI video. It decides what gets generated, when, and how. The decentralization maximalists will scream. But the market doesn’t care. Efficiency wins.
The crash wasn’t the market; it was the architecture. The crash of fragmented workflows is what Preview solves. The architecture of a single, coherent metadata layer is more valuable than decentralization for its own sake.
However, there’s a deeper contrarian angle: Preview’s metadata layer is actually a perfect foundation for blockchain-based verification. Each frame’s provenance could be hashed and anchored to a public ledger. That would solve the deepfake detection problem without requiring a centralized authority. Preview could become the gateway to on-chain video authenticity.
Governance isn’t a democracy; it’s leverage waiting to be wielded. Preview’s governance of its metadata schema will determine the future of AI video standards. If they open-source the format, they become the industry standard. If they keep it proprietary, they become a walled garden. The leverage is in their hands.
Takeaway: The Next Watch
Preview’s true test will be when the first dispute arises over AI-generated content ownership. Will a studio be able to produce the metadata to prove provenance? If yes, Preview becomes indispensable. If no, the whole system collapses. That’s the moment to watch.
Speed is the only currency that doesn’t lose value. Preview raised $12M in six months, but the real race is against time — the time until a court case demands a frame’s digital birth certificate. I’ll be watching the metadata, not the valuation.
