Sign in Subscribe
Western Hemisphere

Sequoia Leads $12M Round for Preview, the Production Platform for AI Video

The Editor

The Editor

— 4 min read
Share
Preview Founders Stefan Fejes and Veljko Tornjanski
Preview Founders Stefan Fejes and Veljko Tornjanski

Preview, a platform that manages professional video production workflows built around generative AI, has raised US$12 million led by Sequoia Capital, with participation from The General Partnership and investor Badrul Farooqi. The announcement came on August 13. The round also drew angel investors including Frame.io co-founder Emery Wells, former Mozilla chief executive John Lilly, Gumroad founder Sahil Lavingia, and designer Soleio.

The company was founded in early 2026 by Stefan Fejes and Veljko Tornjanski, friends since high school in Serbia who were both on the founding team at video editing company Veed, and later worked together at Vizcom, where Tornjanski was first hire and head of AI while Fejes led product. Preview combines a video timeline with an infinite canvas in a collaborative workspace, handling script breakdown into scenes and characters, shot lists, side-by-side generation across multiple frontier models, 3D navigable sets for finding camera angles, and asset libraries that keep characters, locations, and props consistent across shots. Its distinguishing feature is provenance: a complete record of the model, settings, and person behind every frame. The platform carries SOC 2 Type II readiness, ISO 27001 certification, and zero data retention on customer content. More than 100 studios use it, including agencies producing commercials for Fortune 100 brands and filmmakers working on hybrid productions with Oscar-winning talent due on streaming platforms later this year, and 3,000 studios sit on the waitlist.

Market Context

Generative video has moved from demonstration to production faster than most creative technologies. The AI video generation market reached roughly US$6.2 billion in 2025 and is forecast to grow at about 25% annually, while 73% of Fortune 500 companies have integrated AI video tools into content production and 63% of video marketers used AI for creation or editing in 2025, up from 51% a year earlier. Production timelines have compressed dramatically, with a 60-second marketing video that once took nearly two weeks now achievable in under half an hour.

Sequoia's thesis is that the models are no longer the constraint. In its investment note the firm argues video models have become capable enough to substitute for cameras in many contexts, leaving human creativity and inference cost as the real bottlenecks, and describes Preview as "a Cursor for video," referencing the AI coding tool. That framing points at the gap the company is filling: professional production still requires scripts, shot lists, version control, reviews, deliveries, and documentation of how each frame was made, none of which a text-to-video model provides. Morgan Stanley estimates generative AI could cut production costs by roughly 30%, which is why studios are adopting despite unresolved questions about how the work gets credited.

The Signal

"AI video is now part of how real productions get made. What professional work requires hasn't changed: scripts, shot lists, versions, reviews, deliveries, and a clear record of how every frame was made." — Stefan Fejes, co-founder, Preview

Regional Relevance

For the United States: Preview is selling into an industry still working through the terms it set after the 2023 strikes. Collective bargaining agreements now require consent and payment before studios create digital replicas of performers, the Alliance of Motion Picture and Television Producers has confirmed AI cannot receive writing or primary creator credit, and the Academy bars AI-generated actors and scripts from Oscar eligibility while reserving the right to ask how AI was used. In that environment, a system that logs which model produced which frame and who directed it is not merely a workflow convenience; it is potentially the compliance layer studios need to demonstrate what was human-made. Nearly 68% of filmmakers say AI use should be disclosed to audiences, and no standard for doing so yet exists.

For Serbia and Central Europe: Preview is a Serbian-founded company backed by Sequoia within months of its founding, an outcome that remains rare for the region. Both founders built their track record inside other European and American startups, Veed and Vizcom, before starting their own, a path that reflects how technical talent from smaller European markets typically reaches this level. For Serbia's technology sector, which has grown substantially as an engineering base for foreign companies, a locally founded company raising from a top-tier Silicon Valley fund is a different category of validation than outsourcing contracts.

The Other Side

Is a workflow layer defensible when the models beneath it keep changing? Preview's value comes from orchestrating across frontier models rather than owning one, which avoids the enormous cost of training but leaves the company dependent on providers that could build comparable production tooling themselves. Adobe acquired Frame.io, whose co-founder is an angel here, precisely because review and collaboration layers are attractive to incumbents, and that dynamic cuts both ways.

Does 3,000 waitlisted studios indicate demand or gated supply? A waitlist is a controllable number, and the company has disclosed no revenue, pricing, or retention data. More than 100 studios in active use is a meaningful figure for a company founded this year, but the gap between the two numbers is a marketing decision as much as a measure of unmet demand.

Who benefits from provenance tracking, and who is exposed by it? A complete record of how each frame was made serves studios facing questions about disclosure and credit, but the same record documents precisely how much of a production was generated rather than filmed. In an industry where unions negotiated hard over digital replicas and where independent filmmakers already argue they lack access to the training data major studios control, better documentation of AI use is a tool whose consequences depend entirely on who is permitted to see it.

Sources & Transparency

Read more