wtf is live-api-web-console?

The Live API Web Console is a starter app that lets you build and experiment with real-time, voice- and video-driven conversations powered by Google's Gemini AI. Instead of typing a prompt and waiting for text to come back, you can talk to the AI through your microphone, show it your webcam or screen, and hear it respond out loud, all in a browser window. It's essentially a ready-to-run template for creating interactive AI assistants that feel more like a live phone or video call than a chatbot. Under the hood, the app maintains a continuous connection (called a websocket) to Gemini's Live API. It handles the messy plumbing for you: capturing audio from your mic, streaming video from your camera or screen, and playing back the AI's spoken responses in real time. It also includes a built-in log view so you can see exactly what data is flowing back and forth, which is invaluable when you're trying to understand or debug what the AI is doing. You just plug in a free Gemini API key, run a couple of commands, and you have a working live AI conversation running locally. This project is aimed at developers and builders who want to create real-time multimodal AI experiences without starting from scratch. The repository includes several example apps on different branches to show what's possible: one demonstrates "proactive audio" where the AI can initiate conversation, another acts as an explainer, and others handle weather or list-making tasks. The code example in the README shows how you could let the AI search Google and then automatically render a data graph in your app based on what it finds. It's worth noting that Google explicitly calls this an experiment rather than an official product, so while it's a solid starting point for prototyping, you should temper expectations around long-term support. The project is built with React and Create React App, making it approachable for anyone familiar with basic web development.