/easy-whisper-transcribe

Primary LanguagePythonMIT LicenseMIT

Easy Whisper Transcribe

Overview

Easy Whisper Transcription is an advanced voice-to-text conversion tool leveraging OpenAI's Whisper models. Designed for efficiency and ease of use, it offers real-time audio transcription with support for various input devices and customization options.

Features

  • Real-time voice transcription using OpenAI's Whisper models.
  • Continuous audio recording with adjustable sensitivity.
  • Dynamic energy threshold for enhanced recognition.
  • Command-line interface for easy interaction.
  • Options to save transcriptions and audio files.

Available Models

The Easy Whisper Transcription tool supports the following models from OpenAI's Whisper series:

  • tiny: 39 million parameters (supports both English-only and Multilingual transcription)
  • base: 74 million parameters (supports both English-only and Multilingual transcription)
  • small: 244 million parameters (supports both English-only and Multilingual transcription)
  • medium: 769 million parameters (supports both English-only and Multilingual transcription)
  • large: 1550 million parameters (supports Multilingual transcription only)
  • large-v2: 1550 million parameters (supports Multilingual transcription only)
  • large-v3: 1550 million parameters (supports Multilingual transcription only)

Software Stack

The Easy Whisper Transcription tool is built upon a robust and efficient software stack:

  • Python: For its simplicity and vast ecosystem.
  • OpenAI's Whisper Models: For high-quality, multilingual voice transcription.
  • PyTorch: For model development and deployment.
  • Speech Recognition: To enhance voice recognition capabilities.
  • PyDub: For handling audio data.
  • NumPy: For efficient data array processing.
  • Rich: To enhance the CLI experience.
  • Google Fire: For automatic CLI generation.
  • Accelerate (by Hugging Face): For optimizing model performance.
  • YAML: For storing transcription outputs.

Installation

Prerequisites

  • Python 3.6 or later.
  • ffmpeg: Install via conda or your system's package manager. For macOS, use brew install ffmpeg.
  • PyAudio: For macOS, first install portaudio using brew install portaudio and then pip install pyaudio. For Linux, install portaudio19-dev first.

Steps

  1. Clone the Repository 
    git clone https://github.com/yourusername/easy-whisper-transcribe.git
cd easy-whisper-transcribe
  2. Setup a virtual env 
    python -m venv venv
source venv/bin/activate  # On Windows use `venv\Scripts\activate`
  3. Install dependencies 
    pip install -r requirements.txt

Usage

Run the script from the command line, providing necessary arguments:

python app.py --model_name base --verbose True --energy 300

Contributing

Contributions to Easy Whisper Transcribe are welcome! Please read our contributing guidelines to get started.

License

This project is licensed under the MIT License.

Acknowledgements

This project utilizes OpenAI's Whisper models for voice transcription. Inspiration was taken from the whisper-mic project that was key to the development of this project.