Real-time speech-to-text for the deaf and hard-of-hearing community.

AuraVision captures microphone audio, runs it through Vosk speech recognition, and displays the text live in your browser. Supports English, Spanish, and Farsi with proper RTL rendering.

• Real-time transcription — text appears as you speak, with interim results
• Multi-language — English, Spanish, Farsi (auto-detected from model)
• RTL support — Farsi text renders right-to-left with Noto Sans Arabic font
• Microphone picker — choose input device, see real-time audio levels
• Dark UI — audio visualizer, transcript history, export to text file
• Docker ready — one-command production setup with gunicorn
• Auto-detect models — drop any Vosk model in models/ and pick it at startup
• Auth token — optional AV_AUTH_TOKEN for network deployments

git clone https://github.com/Rfannn/AuraVision.git
cd AuraVision

# Linux / macOS
chmod +x init.sh
./init.sh

# Windows
init.bat

Grab Vosk models from alphacephei.com/vosk/models and extract into models/:

curl -O https://alphacephei.com/vosk/models/vosk-model-small-fa-0.5.zip
unzip vosk-model-small-fa-0.5.zip -d models/

Any Vosk model works — the app scans models/ and lets you choose.

Two terminals needed:

# Terminal 1 — Web server
source venv/bin/activate   # Windows: venv\Scripts\activate
python app.py

# Terminal 2 — Speech recognition
source venv/bin/activate
python main.py

Open http://localhost:5000. Select your microphone from the dropdown and start speaking.

docker compose up --build

The web server runs via gunicorn in the container. main.py must run on the host (needs microphone).

Copy .env.example to .env:

Browser ◀──WebSocket── app.py (Flask + SocketIO) ◀──HTTP POST── main.py ──▶ PyAudio ──▶ Vosk

• main.py — captures microphone audio, runs Vosk recognition, POSTs text to app.py
• app.py — receives text, pushes it to all connected browsers via Socket.IO
• config.py — shared configuration (paths, audio settings, Flask config, auth)
• static/style.css — styles
• static/app.js — frontend logic (mic selection, audio levels, transcript)
• templates/index.html — minimal HTML shell

AuraVision/
├── app.py              # Flask + SocketIO web server
├── main.py             # Microphone + Vosk speech recognition
├── config.py           # Centralized configuration
├── gunicorn.conf.py    # Production WSGI config
├── requirements.txt    # Python dependencies
├── Dockerfile          # Container image (gunicorn)
├── docker-compose.yml  # Docker Compose config
├── static/
│   ├── style.css       # Styles
│   ├── app.js          # Frontend logic
│   └── favicon.svg     # App icon
├── templates/
│   └── index.html      # HTML shell
├── tests/
│   └── test_app.py     # pytest tests
├── models/             # Vosk models (gitignored)
├── .github/
│   ├── workflows/ci.yml
│   ├── ISSUE_TEMPLATE/
│   ├── PULL_REQUEST_TEMPLATE.md
│   └── CODEOWNERS
├── CHANGELOG.md
├── CONTRIBUTING.md
├── ROADMAP.md
├── SECURITY.md
└── LICENSE

See CONTRIBUTING.md for development setup and guidelines.

See ROADMAP.md for planned features.

See CHANGELOG.md for release history.

MIT — see LICENSE.

• Email: tsmrfangg@gmail.com
• Telegram: @GNS_Rfan
• Discord: gnsrfan