Partnering with Preview: Lights, Inference, Action

Sequoia is leading Preview’s seed round, betting that the next big AI workflow won’t be code. It’ll be video. The firm’s case is simple: video models have moved fast enough that the bottleneck is no longer making the machine understand motion. The bottleneck is the messy human side of production, where files are huge, approvals are scattered, and the paper trail is often a headache.

Preview is trying to replace that sprawl with one AI-native workspace. The platform combines a video timeline with an infinite canvas for brainstorming, and it lets teams generate with any model while keeping track of every character, location, and prop. It also keeps records of how each asset was made, which matters when a finished shot needs to be traced back for legal clearance. That is the kind of unglamorous plumbing that usually decides whether a creative tool becomes part of the job or stays a demo.

The pitch is aimed at professional creators, not hobbyists. Sequoia says more than 100 studios are already using Preview, including agencies making commercials for Fortune 500 brands, AI-native studios, and Hollywood feature filmmakers. Customers, according to the post, see it as the most important tool in their workflow and the central source of truth for what happens next.

The founders have the backstory Sequoia likes to tell. Stefan Fejes and Veljko Tornjanski grew up together in Serbia, were both early at Veed, and later split roles at Vizcom, where Veljko was the first hire and became head of AI while Stefan led product. Sequoia’s larger bet is that video is heading into a period of heavy inference spend, and the platform where creative work actually gets done will capture the value. That is a neat thesis, and one that says more about infrastructure than inspiration.