Speech2Text - Audio Transcription

Advanced speech-to-text transcriber using HuggingFace transformers and Facebook's Hubert model for accurate audio transcription.

🎯 What This Project Does

• 🎙️ Multi-Format Support - Process MP3, MP4, WAV audio and video files
• 🧠 AI-Powered Transcription - Uses Facebook's Hubert model for accurate speech recognition
• 📝 Downloadable Transcripts - Get clean, formatted text output
• ⏱️ Timestamped Segments - View transcription with time markers

🌐 Live Demo

📺 Demo Video

🚀 How to Use

Using the App

1. Upload Audio/Video - Support for .mp3, .mp4, .wav formats
2. Watch Processing - Real-time transcription progress
3. View Results - Timestamped segments and full transcript
4. Download Text - Export transcript in clean text format

Local Development

# Clone and setup
git clone https://github.com/aishwaryaj7/speech2text.git
cd speech2text

# Install dependencies
pip install -r requirements.txt

# Run the app
streamlit run src/app.py

🔧 Key Features

• Multi-Format Support: MP3, MP4, WAV audio and video files
• AI-Powered Transcription: Facebook's Hubert model for accurate speech recognition
• Timestamped Segments: Precise time markers for each phrase
• Real-time Processing: Live transcription with progress indicators
• Download Options: Export transcripts in multiple formats
• User-Friendly Interface: Intuitive Streamlit interface

🛠️ Tech Stack

AI/ML: HuggingFace Transformers, Hubert Model, Wav2Vec2 Audio: Torchaudio, Librosa, Pydub Frontend: Streamlit Deployment: Streamlit Cloud

🤝 Skills Demonstrated

• Speech Recognition & Audio Processing
• HuggingFace Transformers & Model Deployment
• Streamlit Development & Cloud Deployment
• Signal Processing & Feature Extraction
• User Experience Design