=====================================================
UNIFIED AI VOICE SERVER (TTS + STT) - WINDOWS RTX GPU
=====================================================

This folder contains the complete unified server designed to run BOTH Text-to-Speech (Kokoro) and Speech-to-Text (Whisper) on your Windows RTX 3060 simultaneously.

--- STEP 1: PREREQUISITES ---
1. You must have Python installed on your Windows PC. 
   - Go to python.org/downloads and download Python 3.10 or 3.11.
   - VERY IMPORTANT: Check the box that says "Add Python to PATH" during installation.

--- STEP 2: START THE SERVER ---
1. Double-click the `start_server.bat` file in this folder.
2. A black command prompt window will open. It will automatically:
   - Create a virtual environment
   - Install FastAPI, WebSockets, Kokoro, and Faster-Whisper.
   - Download the Kokoro AI models and the Faster-Whisper base.en model.
   - Start the server on Port 5000.
3. Wait until it says "Application startup complete."

--- STEP 3: EXPOSE TO THE INTERNET WITH NGROK ---
Because Vapi needs to stream audio over WebSockets to your PC, you need a secure tunnel.

1. Go to https://ngrok.com/download and download the Windows version.
2. Unzip it and double-click `ngrok.exe`.
3. In the new Ngrok window, type:
   ngrok http 5000
4. Press Enter. 
5. Ngrok will show a "Forwarding" URL (e.g., https://a1b2c3d4.ngrok-free.app)

--- STEP 4: UPDATE VAPI ---

FOR TEXT-TO-SPEECH (Custom Voice):
1. In your Vapi Dashboard, edit your Voice Provider to "Custom".
2. Set the Server URL to: https://[your-ngrok-url].ngrok-free.app/vapi/tts

FOR SPEECH-TO-TEXT (Custom Transcriber):
1. In your Vapi Dashboard, edit your Transcriber to "Custom".
2. Set the Server URL to your Ngrok websocket URL. IMPORTANT: You must change "https://" to "wss://"
   Example: wss://[your-ngrok-url].ngrok-free.app/vapi/stt

Your RTX 3060 is now powering BOTH the hearing (STT) and speaking (TTS) of your real estate agent in real time!