Hands-Free AI Voice Assistant for Windows (Wake Word to Reply)
Typing is fine when you're at the keyboard. But a lot of the time you want to just ask — while reading code, cooking, or walking across the room. A real hands-free AI voice assistantlets you say a wake word, speak naturally, and hear the answer back, without touching anything. Here's how it works on the Windows desktop, and how to turn it on.
Wake word: it's listening for you, not recording you
You start a conversation by saying "Hey Jarvis." The wake word is detected locally on your machine — the assistant isn't streaming your room to the cloud, it's watching for one phrase and ignoring everything else until it hears it. When it triggers, it starts listening for your actual request.
From speech to spoken answer
Under the hood, a voice turn runs as a short pipeline, and each stage is chosen for latency and privacy:
- Speech to text — your words are transcribed by Whisper (running locally with Faster-Whisper) or an optional premium cloud engine.
- The agent thinks — the transcript goes to the same agent that powers everything else, so it can actually act, not just answer.
- Text to speech — the reply is spoken back with a natural local voice (Kokoro) or a premium cloud voice if you've added one.
Replies stream sentence by sentence, so it starts talking before the whole answer is generated — the felt latency is a beat, not an awkward wait.
Real conversation, not one-shot commands
After it answers, it can keep listening for your follow-up without making you say the wake word again — so "what about tomorrow?" just works. When enabled, it can also check in proactively during idle moments, and it politely stays quiet during your configured quiet hours so it never pipes up at 2am.
It's a genuine back-and-forth: wake word → your question → spoken answer → your follow-up. And because it's the full assistant behind it, "turn off the music" or "summarize my unread email" actually happen — voice is just another way in.
Turning it on (and the Windows mic bit)
Enable voice in Settings → Voice, then turn on hands-free mode. One Windows-specific step matters: make sure desktop apps are allowed to use your microphone under Windows Settings → Privacy → Microphone. Without that OS permission, no app can hear you — it's the most common reason voice "does nothing" on a fresh install.
Local or cloud — your call
Out of the box, speech recognition and the voice both run locally, so hands-free works offline and private. If you want the most natural-sounding voice, you can opt into a premium cloud engine with an API key — but it's never required, and it falls back to local automatically if the cloud is unavailable.
Hands-free voice turns your desktop assistant into something you talk to, not just type at. Wake word in, spoken answer out, private by default.
Stop reading. Start commanding.
Jarvis is free. Install it, add a key (or run local), and give it a real task.