Arabic Speech Recognition

A browser-based Arabic speech recognition app powered by OpenAI Whisper. Upload a WAV file and get an accurate Arabic transcription instantly — no training, no setup, no API keys required.

---

Features

• Automatic transcription of Arabic speech from WAV audio files
• Pre-trained Whisper model — works out of the box on first run
• Waveform preview rendered in the browser before transcription
• Right-to-left Arabic output display
• Clean, responsive web interface

---

How It Works

At startup the app loads OpenAI's whisper-base model (downloaded automatically on first run, ~140 MB). When you upload a WAV file, Whisper processes it with the Arabic language hint and returns the transcription as Arabic text.

---

Project Structure

├── web_app.py           Flask web server — main entry point (port 5000)
├── requirements.txt     Python dependencies
├── README.md            This file
├── .gitignore
├── .replit
└── scripts/
    └── post-merge.sh    Post-merge dependency installer

---

Getting Started

Install dependencies

pip install -r requirements.txt

Run the server

python web_app.py

Open http://localhost:5000 in your browser. The status banner turns green once the model is ready. On the first run, Whisper downloads the model weights automatically (~140 MB).

---

Usage

1. Click the upload area or drag and drop a WAV file onto it.
2. A waveform preview is drawn automatically.
3. Click Transcribe.
4. The Arabic transcription appears in the result box (right-to-left).

Supported audio: WAV format, mono or stereo, any sample rate (Whisper resamples internally). Files up to 50 MB are accepted.

---

Dependencies

Package	Purpose
openai-whisper	Pre-trained multilingual ASR model
flask	Web framework and HTTP server
numpy	Numerical array operations

---

License

Project code is released under the MIT License. The Whisper model weights are released by OpenAI under the MIT License.