whisper-cli

Local speech-to-text CLI tool using faster-whisper.

This work was inspired and forked from whisper.cpp. For the best performance on Mac with Apple Silicon, please use the original work instead.

Say Goodbye to Transcription Limits and Privacy Risks.

Most online transcription services come with a catch: they only allow a few minutes for free, force you through tedious registration, or hide their best features behind expensive subscriptions.

whisper-cli is here to change that. Powered by OpenAI’s world-class "Whisper" model—specifically the high-performance faster-whisper implementation—you can now transcribe audio in over 99 languages completely locally on your own machine.

• Zero Cost: No subscriptions, no hidden fees, no limits.
• Total Privacy: Your data never leaves your device. Once the model is downloaded, you can even use it entirely offline.
• Uncompromising Speed: Optimized to run efficiently on your hardware, whether you have a high-end GPU or a standard CPU.

Note: The script handles the model download for you automatically during the first run. No manual downloading required!

Features

• Fully offline transcription
• OpenAI Whisper (faster-whisper) technology
• Support for 99+ languages
• TXT / SRT output
• Timestamp support
• Beginner-friendly CLI

Prerequisites

Before installing, ensure you have Python and pip installed on your system.

Installing Python and pip

• Windows: Download and run the installer from the official Python website. Make sure to check the box "Add Python to PATH" during installation.
• macOS: Python 3 and pip3 are usually pre-installed or can be installed via Homebrew: brew install python.
• Linux: Use your distribution's package manager.
  • Ubuntu/Debian: sudo apt update && sudo apt install python3 python3-pip
  • Fedora: sudo dnf install python3 python3-pip

Installation

Follow these steps to set up the tool on your local machine:

1. Clone the Repository

   git clone https://github.com/yuutakun-2/audio-transcriber.git
   cd audio-transcriber

2. (Optional) Create a Virtual Environment It is recommended to use a virtual environment to avoid dependency conflicts.

   # Windows
   python -m venv venv
   venv\Scripts\activate
   
   # macOS/Linux
   python3 -m venv venv
   source venv/bin/activate

3. Install the Package

   On Windows and Linux:

   pip install .

   On macOS: Use pip3 to ensure the tool and its dependency faster-whisper are installed correctly:

   pip3 install .

Usage

Run the tool by typing the command in your terminal:

whisper-cli

The tool is interactive and will prompt you for the following settings:

Configuration Options:

• Model: Choose the model size (tiny, base, small, medium, large). Larger models are more accurate but slower and require more memory.
• Language: Specify the language (e.g., ja for Japanese, en for English) or use auto for automatic detection.
• Device: Select cpu or auto. Use auto if you have a compatible NVIDIA GPU (CUDA) for much faster processing.
• Compute: Selection of precision (int8, float16, float32). Default int8 is recommended for most CPUs.
• Beam size: Number of beams for search (1-5). Higher values might improve accuracy slightly.
• Output format:
  1. TXT with timestamps: [00:00:10 - 00:00:20] Transcription text
  2. TXT only: Plain text output.
  3. SRT: Subtitle format.
• Skip existing: If y, the tool will skip files that already have a corresponding output file in the destination folder.
• Input path: Provide an absolute or relative path to an audio file or a folder containing multiple audio files.
• Output folder: Specify where to save the results. Defaults to the same location as the input.
• Save settings: If y, your choices will be saved to ~/.whisper-cli/config.json and used as defaults for the next run.