Document & Audio Processing Service

A comprehensive web-based application that provides three powerful features:

• Audio Transcription: Convert audio files to text using OpenAI's Whisper AI
• Document Conversion: Convert documents between various formats (DOCX, PDF, TXT, Excel)
• OCR (Optical Character Recognition): Extract text from images and PDF files

• Multiple Audio Format Support: MP3, WAV, M4A, FLAC, OGG, OPUS, AAC, WMA, MP4, WebM, 3GP, AMR, AIFF, AU
• Automatic Language Detection: Automatically detects the language in your audio
• High-Quality Transcription: Powered by OpenAI's Whisper model (base model by default)
• Translation Support: Translate transcriptions to English, French, Spanish, German, or Dutch
• Real-time Progress Tracking: Visual feedback with elapsed time and processing steps
• Sentence Formatting: Automatically formats transcriptions with sentences on separate lines
• Dual Output: View both original and translated text side-by-side

• DOCX Conversions: DOCX ↔ TXT, DOCX ↔ PDF
• PDF Conversions: PDF ↔ TXT, PDF ↔ DOCX
• TXT Conversions: TXT ↔ PDF, TXT ↔ DOCX
• Excel Conversions: XLSX/XLS → TXT
• Format Detection: Automatically detects available conversion options based on file type

• Image Support: PNG, JPG, JPEG, GIF, BMP, TIFF, WEBP
• PDF Support: Extract text from PDF files (multi-page support)
• Formatted Output: Extracted text formatted with sentences on separate lines
• Page Separation: PDF pages are clearly marked in the output

• Modern Tabbed Interface: Easy navigation between features
• Drag & Drop Support: Intuitive file upload
• Real-time Progress: Live timers and progress bars
• Processing Time Display: See exactly how long operations take
• Download Results: Download all processed files

• Python: 3.8 or higher
• FFmpeg: Required for audio transcription
• Tesseract OCR: Required for OCR functionality (optional but recommended)

Windows:

# Using winget (recommended)
winget install ffmpeg

# Or using Chocolatey
choco install ffmpeg

# Or download from: https://ffmpeg.org/download.html

macOS:

brew install ffmpeg

Linux (Ubuntu/Debian):

sudo apt update
sudo apt install ffmpeg

Windows:

# Using winget (recommended)
winget install --id UB-Mannheim.TesseractOCR --accept-package-agreements --accept-source-agreements

macOS:

brew install tesseract

Linux (Ubuntu/Debian):

sudo apt update
sudo apt install tesseract-ocr

1. Clone or navigate to the project directory:

cd audiotranscribe

2. Create a virtual environment (recommended):

python -m venv venv

3. Activate the virtual environment:

   Windows:

   venv\Scripts\activate

   macOS/Linux:

   source venv/bin/activate

4. Install dependencies:

pip install -r requirements.txt

5. Download Whisper model (automatic on first use): The application will automatically download the Whisper model on first transcription. The "base" model is used by default.

See DOCKER.md for detailed Docker installation instructions.

Quick start:

docker-compose up -d

Windows:

python app.py
# Or double-click start.bat

macOS/Linux:

python app.py
# Or
bash start.sh

Docker:

docker-compose up -d

Open your web browser and navigate to:

http://localhost:5012

1. Click the "🎤 Audio Transcription" tab
2. Drag and drop an audio file or click to browse
3. (Optional) Select a target language for translation (English, French, Spanish, German, or Dutch)
4. Click "Transcribe Audio"
5. Wait for processing (you'll see real-time progress)
6. View and download the transcription (and translation if requested)

1. Click the "📄 Document Conversion" tab
2. Upload a document (DOCX, PDF, TXT, or Excel file)
3. Select the target format from the dropdown
4. Click "Convert Document"
5. Download the converted file

1. Click the "👁️ OCR (Text Extraction)" tab
2. Upload an image or PDF file
3. Click "Extract Text (OCR)"
4. View and download the extracted text

• MP3, WAV, M4A, FLAC, OGG, OPUS, AAC
• WMA, MP4 (audio track), WebM (audio track)
• 3GP, AMR, AIFF, AU

• Input: DOCX, PDF, TXT, XLSX, XLS
• Output: DOCX, PDF, TXT

• PNG, JPG, JPEG, GIF, BMP, TIFF, WEBP
• PDF (for text extraction)

Edit app.py:

model = whisper.load_model("base")  # Options: tiny, base, small, medium, large

Model Options:

• tiny: Fastest, least accurate (~39M parameters)
• base: Good balance (default, ~74M parameters)
• small: Better accuracy (~244M parameters)
• medium: High accuracy (~769M parameters)
• large: Best accuracy (~1550M parameters)

Default maximum file size is 100MB. To change it, modify MAX_FILE_SIZE in app.py:

MAX_FILE_SIZE = 100 * 1024 * 1024  # 100MB

Default port is 5012. To change it, modify the last line in app.py:

app.run(debug=True, host='0.0.0.0', port=5012)

audiotranscribe/
├── app.py                    # Flask backend server
├── transcribe_file.py        # Command-line transcription script
├── requirements.txt          # Python dependencies
├── Dockerfile               # Docker container configuration
├── docker-compose.yml       # Docker Compose configuration
├── .dockerignore           # Docker ignore patterns
├── templates/
│   └── index.html          # Frontend web interface
├── uploads/                # Temporary upload directory (auto-created)
├── transcriptions/          # Saved transcription files (auto-created)
├── conversions/             # Converted document files (auto-created)
├── ocr_results/            # OCR extracted text files (auto-created)
├── start.bat               # Windows startup script
├── start.sh                # Linux/macOS startup script
├── README.md               # This file
├── DOCKER.md               # Docker documentation
└── INSTALL_FFMPEG.md       # FFmpeg installation guide

• GET / - Main web interface
• POST /upload - Upload and transcribe audio file (supports target_language parameter)
• GET /download/<filename> - Download transcription file
• GET /supported-formats - Get list of supported audio formats
• GET /check-ffmpeg - Check FFmpeg installation status
• GET /translation-capabilities - Get translation capabilities and supported languages

• POST /convert-document - Convert document between formats
• GET /download-conversion/<filename> - Download converted document
• GET /supported-conversions - Get supported conversion formats

• POST /ocr - Perform OCR on image or PDF
• GET /download-ocr/<filename> - Download OCR result
• GET /ocr-capabilities - Get OCR capabilities and status

• GET /health - Health check endpoint

See DOCKER.md for complete Docker deployment instructions.

# Build and start
docker-compose up -d

# View logs
docker-compose logs -f

# Stop
docker-compose down

• Ensure FFmpeg is installed and available in your system PATH
• Verify installation: ffmpeg -version
• See INSTALL_FFMPEG.md for detailed instructions

• Install Tesseract OCR (see Prerequisites section)
• On Windows, the app will auto-detect Tesseract in common locations
• Verify installation: tesseract --version
• Restart the server after installation

• Ensure all dependencies are installed: pip install -r requirements.txt
• Restart the server after installing new packages
• Check that Python version is 3.8 or higher

• Use a smaller Whisper model (tiny or base) for faster processing
• Consider using GPU acceleration if available (requires CUDA-enabled PyTorch)
• Processing time is roughly 0.2-0.5x the audio duration

• Use a smaller Whisper model
• Process shorter audio files
• Close other applications to free up memory
• Reduce MAX_FILE_SIZE if processing large files

• Ensure you have a stable internet connection for the first run
• Models are cached after first download
• First transcription may take longer as the model downloads

• Verify that required libraries are installed (python-docx, pypdf, reportlab, etc.)
• Check server logs for specific error messages
• Ensure file format is supported (see Supported Formats section)

1. GPU Acceleration: For faster transcription, install PyTorch with CUDA support:

   pip install torch torchaudio --index-url https://download.pytorch.org/whl/cu118

2. Model Selection: Choose the appropriate model size based on your needs:

   • Quick transcriptions: tiny or base
   • High accuracy: medium or large

3. File Preparation: For best results:

   • Use clear audio with minimal background noise
   • Ensure good audio quality
   • Consider splitting very long files
   • For OCR, use high-resolution images

4. Docker Performance:

   • Use Docker volumes for persistent storage
   • Allocate sufficient memory (recommended: 4GB+)
   • Consider using GPU support in Docker for faster processing

• Local Processing: All processing happens locally - no data is sent to external services
• Automatic Cleanup: Uploaded files are automatically deleted after processing
• No Data Storage: Processed files are saved only if you download them
• Secure: All operations run on your local machine or Docker container

• First transcription may take longer as the Whisper model downloads (~150MB for base model)
• Transcription time depends on audio length and model size (roughly 0.2-0.5x audio duration)
• Model loading time: 10-30 seconds on first use, then cached in memory
• Document conversions are typically very fast (< 5 seconds for most files)
• OCR processing time depends on image/PDF size and complexity

This project is licensed under the MIT License - see the LICENSE file for details.

This software uses the following open-source components:

• OpenAI Whisper: MIT License - Copyright (c) 2022 OpenAI
• Tesseract OCR: Apache License 2.0 - Copyright (c) Google Inc.
• Other dependencies: See requirements.txt for individual licenses

All third-party components maintain their original licenses. Please refer to the LICENSE file for complete license information and third-party attributions.

Feel free to submit issues, fork the repository, and create pull requests for any improvements.

For issues or questions:

1. Check the Troubleshooting section above
2. Review the documentation files (DOCKER.md, INSTALL_FFMPEG.md)
3. Check server logs for detailed error messages
4. Open an issue on the repository

• Content Creators: Transcribe podcasts, videos, and interviews
• Students: Convert lecture recordings to text notes
• Business: Extract text from scanned documents and PDFs
• Accessibility: Make audio content accessible through text
• Document Management: Convert documents between formats
• Data Entry: Extract text from images and forms