anirxdh/voice-companion

Super Nova is a voice-first AI companion web app where you control everything by speaking, no typing required. You tap an orb to activate it, then ask for weather, directions, music, news, NASA photos, Wikipedia summaries, YouTube videos, timers, or general questions. Each request produces an animated widget on screen and a spoken response that streams back through ElevenLabs text-to-speech, so the voice reply starts before the full answer is even generated. The animated environment includes a pet character that walks between different themed "rooms" depending on what you asked for, a music room, a map room, a weather room, and so on, giving the interface a spatial feel. Under the hood, your speech is captured by the browser's built-in speech recognition API, then sent to the Groq LLM (which runs fast intent classification to understand what you're asking for in about 200 milliseconds), routed to the appropriate data source or API, and rendered as a typed widget while ElevenLabs streams the spoken reply. Multiple widgets, weather, a timer, a music player, can all be on screen at the same time without overlapping. The project is built with TypeScript using a web application framework, with three-dimensional rendering for the colony environment and real-time audio visualisation for the microphone and music playback. It requires a Google Chrome browser (the speech recognition API is only available in Chromium), along with API keys for Groq and ElevenLabs, both of which have free tiers. The full README is longer than what was provided.