🎧 Speaker Diarization and Transcription System (Flask + React)

This project consists of:

• Flask server (Python) that receives audio files, converts and transcribes them using VOSK, and estimates the number of speakers.
• React client that lets users upload audio files and view transcription and speaker count.

---

🗂️ Folder Structure

project-root/
├── server/
│   ├── app.py
│   ├── requirements.txt
│   └── models/
│       ├── vosk-model-en-us-0.22/
│       └── vosk-model-spk-0.4/
├── client/
│   ├── package.json
│   ├── public/
│   └── src/
└── README.md

🔧 Setup & Run Instructions

1. Flask Server Setup

Requirements:

• Python 3.8+
• FFmpeg installed (Download: https://ffmpeg.org/download.html)
  Make sure ffmpeg.exe and ffprobe.exe are located at:
  C:\ffmpeg\ffmpeg\bin\

Install dependencies:

cd audio-server
pip install -r requirements.txt

Run the server:

Download VOSK Models:

• Download from: https://alphacephei.com/vosk/models
• Place models inside server/models/:

  vosk-model-en-us-0.22

  vosk-model-spk-0.4

Run server:

python app.py

Default server address: http://localhost:5000

2. React Client Setup

Requirements: Node.js 14+

Install and run:

cd audio-client
npm install
npm start

Client will run on http://localhost:3000.

Make sure the React app sends API requests to http://localhost:5000/analyze, or use a proxy in package.json.

📤 API - Upload Audio File

• Endpoint: POST /analyze
• Content-Type: multipart/form-data
• Required field: file (MP3 or WAV file)

Response:

{
  "status": "success",
  "transcription": "hello this is a test",
  "num_speakers": 2,
  "duration": 9.23
}

📋 Notes

• Audio is automatically converted to mono and 16kHz.
• If speaker model is not found, the response will default to 1 speaker.
• Temporary WAV files are deleted after processing.

---