Generative Video as a Computing Primitive

Video models already make photorealistic clips that are often hard to tell from real footage. YC wants founders who treat generated video as a basic building block for software, not just an output.

Google's Veo 3 already produces 8-second, photorealistic clips with sound for a few dollars each, often indistinguishable from reality. YC expects that soon anyone will be able to generate near-perfect footage of anything for a marginal cost approaching zero, turning video into a new basic building block for software.

Where it applies:

• •Media: a new season of a canceled show, a cartoon starring your own family, a TikTok successor where every video is made for one viewer
• •Shopping: seeing yourself in the clothes you browse, or your own things auto-staged in apartment listings
• •Gaming and simulation: games with no game engine, APIs that return endless robot training data, video calls with loved ones after they're gone

The ask: new apps, tooling and infrastructure for a world of limitless, low-latency video.

Read the full request on YC →