🎤 Whisper AI Transcription App

A modern, user-friendly Streamlit web interface for OpenAI's Whisper speech-to-text model, containerized with Docker for easy deployment.

✨ Features

• 🖱️ Drag & Drop Interface: Simply drag audio files into the browser or click to browse
• 🧠 Model Selection: Choose from 5 different Whisper models based on your accuracy/speed needs:
  • tiny.en - Fastest, least accurate (~39MB)
  • base.en - Good balance (~74MB)
  • small.en - Better accuracy (~244MB)
  • medium.en - High accuracy (~769MB)
  • large - Best accuracy (~1550MB)
• 📁 Flexible Output: Save files to default location or specify custom paths
• 📥 Download Buttons: Download transcript files (TXT, SRT, VTT) directly from the web interface
• 🐳 Docker Volume Support: Map container directories to host machine for persistent file access
• 💾 Memory Optimized: CPU-only processing with memory-friendly options
• 📱 Responsive UI: Clean, modern interface that works on desktop and mobile

🚀 Quick Start

Build the Docker Image

git clone https://github.com/jlonge4/whisperAI-flask-docker.git
cd whisperAI-flask-docker/whisperapp
docker build -t whisper-streamlit .

Run the Container

Basic usage (download files via web interface):

docker run -p 8501:8501 whisper-streamlit

With volume mapping (optional - for direct file access):

docker run -p 8501:8501 -v ~/transcripts:./whisper_files whisper-streamlit

With extra memory (for larger models):

docker run --memory=4g -p 8501:8501 -v ~/transcripts:./whisper_files whisper-streamlit

Access the App

Open your browser and go to: http://localhost:8501

🎯 How to Use

1. Upload Audio: Drag and drop your audio file or click to browse
   • Supported formats: MP3, WAV, MP4, M4A, FLAC, OGG, WMA
2. Select Model: Choose the Whisper model that fits your needs
3. Choose Output Location: Use default ./whisper_files or specify custom path
4. Start Transcription: Click the transcription button and wait for processing
5. Get Results: Download files instantly using the download buttons (or access via volume mapping if configured)

📂 Output Files

The app generates three transcript formats:

• .txt - Plain text transcript
• .srt - SubRip subtitle format with timestamps
• .vtt - WebVTT subtitle format

🐳 Docker Volume Mapping (Optional)

Volume mapping is completely optional thanks to the built-in download buttons! But if you want direct file system access:

# Map container's ./whisper_files to host's ~/transcripts
docker run -p 8501:8501 -v ~/transcripts:./whisper_files whisper-streamlit

# Map to any host directory
docker run -p 8501:8501 -v /path/to/your/folder:./whisper_files whisper-streamlit

💡 Tips

• Start with smaller models (tiny.en or base.en) to test functionality
• Use larger models (medium.en or large) for better accuracy on complex audio
• Increase Docker memory if you get out-of-memory errors with larger models
• Download buttons provide instant file access without any setup required
• Volume mapping is optional - only needed if you want files to persist on your host machine

🔧 System Requirements

• Docker Desktop (with sufficient memory allocation)
• At least 2GB RAM (4GB+ recommended for larger models)
• ARM64 or AMD64 architecture support

🆕 What's New (Streamlit Version)

This project has been upgraded from Flask to Streamlit with major improvements:

• Modern drag-and-drop interface
• Real-time model selection
• Built-in download functionality
• Better error handling and user feedback
• Memory optimization for Docker containers
• Multi-platform Docker support (ARM64/AMD64)

---

Original concept: A user-friendly way to upload files to a dockerized Flask web form and have Whisper transcribe them. Now with a modern Streamlit interface! 🎉