Awaaz Setu is a web‑based offline speech‑to‑text system that converts spoken audio into accurate text using the Whisper deep learning model. Users can record speech directly from the browser or upload audio files, and the system processes the input to generate real‑time transcriptions without requiring an internet connection.
Awaaz Setu is an offline speech‑to‑text system that converts spoken audio into text using deep learning–based speech recognition.
- 🎤 Records speech directly from the microphone or accepts uploaded audio files
- 🧠 Uses an AI speech recognition model to process audio input
- 📝 Converts spoken language into accurate text transcription
- 🌐 Works completely offline without requiring internet connectivity
- 💻 Provides an interactive web interface for recording, uploading, and viewing results
- 🎤 Live microphone recording
- 📁 Upload audio files (WAV, MP3)
- 🌊 Waveform Visualization
- 🧠 Offline AI Model
- 📄 Transcription Download
- 🌙 Dark Mode Interface
- 🎧 Audio Playback
- 🗂 Transcription History
- React.js
- CSS
- WaveSurfer.js
- Axios
- Node.js
- Express.js
- Multer
- File System (JSON)
- OpenAI Whisper Model
- Python
- Deep Learning Speech Recogniton
- Mozilla Common Voice
- 📁 Upload an audio file or record speech using the microphone
- 🎤 The system captures the audio input from the user
- 🧠 The backend processes the audio using the Whisper speech recognition model
- 📝 The system converts the speech into text transcription
- 📊 The transcription result is displayed on the web interface
- 💾 The transcription can be downloaded or stored for later use
The goal of Awaaz Setu is to develop an offline speech‑to‑text system that converts spoken audio into accurate text using a deep learning–based speech recognition model.
- 🎤 Capture speech through microphone recording or audio file upload
- 🧠 Process audio using the Whisper deep learning model
- 📝 Generate accurate text transcriptions from spoken audio
- 🌐 Enable speech recognition without requiring internet connectivity
- 💻 Provide a simple and user‑friendly web interface for interaction
- 🧑💻 Speaker identification for multi-speaker environments
- 🌍 Improved accent robustness for diverse speech patterns
- 📊 Accuracy evaluation using Word Error Rate (WER) metrics
- 📱 Mobile application integration for on-the-go usage
- 🌐 Multilingual speech recognition support
This project is developed for educational and research purposes.
