We turned ChatGPT into a real voice agent

👀 Here's what happened

Some founders wanted to build a voice agent. They estimated:

• ⏱ 6 months
• 👨‍💻 3 devs
• ☁️ Cloud infrastructure

We said "nah." And built a working prototype in 2 hours – without writing a single line of code.

Just ChatGPT Desktop + Jitsi + Voicemeeter + curiosity.

🎥 Demo (Split Screen)

This is not a mockup – it actually worked.

🔊 Join the Live Room

The voice agent may or may not be active right now. The room is always open. Try speaking. If the ghost responds – it's real ChatGPT.

🛠️ Build It Yourself

1. Install Voicemeeter Potato (free)
2. Install ChatGPT Desktop App and log into ChatGPT Plus
3. Enable Advanced Voice Mode in ChatGPT settings
4. Set Windows default input/output devices for ChatGPT and Jitsi (see below)
5. Launch a free room on Jitsi as the operator (use Edge browser)
6. Send the Jitsi room link to the guest user or embed it using iframe like this:

   <iframe
     class="w-full h-full"
     src="https://meet.jit.si/YOUR_ROOM_NAME#lang=en&config.startWithVideoMuted=true&config.disableVideo=true&config.startWithAudioMuted=false&config.prejoinPageEnabled=false"
     allow="microphone; fullscreen; display-capture"
     allowfullscreen
     sandbox="allow-same-origin allow-scripts allow-popups allow-forms" />

7. Open ChatGPT app and Jitsi room on the same machine
8. Use Voicemeeter to route audio as shown below

Audio Routing Diagram:

VAIO AUX IN ← Jitsi mic (User)
B1 → ChatGPT hears user
ChatGPT → VAIO3
B2 → Jitsi hears ChatGPT
A1 → Operator local monitor

Windows Audio Settings:

Output: Voicemeeter In 3 (VAIO3)
Input: Voicemeeter Out B1
Jitsi mic: Voicemeeter Out B2
ChatGPT mic: Voicemeeter Out B1