Unlocking Client-Side AI in Your Web App

During dotJS 2026 in Paris 🇫🇷 I delivered this talk about running generative AI entirely on the client side

You don't need a cloud subscription or backend infrastructure to bring powerful AI features into your web app. Thanks to modern browser APIs and local model support, it's now possible to run generative AI — including text, vision, and speech — entirely on the client side.

In this session, you'll learn how to unlock AI capabilities in the browser using built-in AI APIs, WebNN, WebGPU, and WebAssembly. We'll explore how on-device models, such as Gemini Nano in Chrome, and open-source models enable local inference for language, vision, and audio: privately, offline, and without recurring cloud costs.

We'll cover architecture patterns, browser and device capabilities, and live demos, including how to detect support and decide when to use built-in models, your own client-side models, or a cloud fallback.

If you're building for the Web, this talk shows how to add real AI to your web app: locally, privately, and using technologies available in the browser today.