Bash Script for Whisper Video Transcription and Translation

Automates subtitle creation for multiple videos using OpenAI's speech recognition model, (Whisper)[https://github.com/openai/whisper].

Overview

This repository provides a bash script tailored to process all video files within a specified directory and its subdirectories. It uses the 'whisper' command, a highly versatile speech recognition model developed by OpenAI, capable of tasks such as multilingual speech recognition, speech translation, and language identification.

The script provides an array of options allowing the user to define the video language, file extensions to be processed, output formats, and the specific task for Whisper. Moreover, it facilitates lowering the CPU and I/O priority of the whisper process using the nice command.

Prior to running this script, please ensure the whisper program is correctly installed on your system. This can be checked by executing the whisper command on its own, for instance:

whisper --task transcribe --language en --model small --output_format txt --output_dir ./ ./myvideo.mp4

If an error message appears or the output doesn't meet expectations, it could signify a problem with your whisper installation.

Installation of Whisper

Before using the script in this repository, install OpenAI's Whisper, a general-purpose speech recognition model. Here are simplified steps to get you started:

1. Python and PyTorch: The Whisper codebase is compatible with Python 3.8-3.11 and recent PyTorch versions. If not already installed, Python can be downloaded here, and PyTorch here.

2. Install Whisper: You can install the latest release of Whisper with the following command:

   pip install -U openai-whisper
   

3. ffmpeg: This tool is necessary for handling video and audio files. Installation instructions vary by operating system:

   • Ubuntu or Debian

     sudo apt update && sudo apt install ffmpeg
     

   • MacOS (using Homebrew)

     brew install ffmpeg
     

For more detailed information, consult the official Whisper documentation.

Installing this Script

After you've installed all the necessary dependencies and have verified that Whisper works correctly on your system, follow these steps to install and use the whisper-batch-subtitler script.

1. Clone the repository: Use the git clone command to clone the whisper-batch-subtitler repository to your local machine:

   git clone https://github.com/mettamatt/whisper-batch-subtitler.git

2. Navigate to the repository folder: After cloning, navigate to the newly created directory:

   cd whisper-batch-subtitler

3. Make the script executable: Before you can run the script, you need to make it executable. You can do this with the chmod command:

   chmod +x whisper-batch-subtitler.sh

Now, you can run the script using the following syntax:

./whisper-batch-subtitler.sh [OPTIONS] <dir>

Please replace <dir> with the directory of the videos you want to process. If you don't specify a directory, the script will use the current one. Refer to the "Usage" section above for information about the different options available with this script.

Remember, always ensure the whisper program is properly installed and functional before running this script.

Usage

./whisper-batch-subtitler.sh [OPTIONS] <dir>

Here, <dir> denotes the directory to process. In the absence of a specified directory, the script will default to the current directory.

Options

• -h, --help Displays help message and exits.
• -e, --extensions <ext> Defines a comma-separated list of file extensions to process. Default is mp4,mkv,avi,flv,wmv.
• -l, --language <lang> Defines the language used in the videos. Default is en.
• -o, --output_format <fmt> Specifies a comma-separated list of output formats. Default is txt,vtt,srt,tsv,json.
• -n, --nice If specified, the script will utilize the 'nice' command to lower the CPU and I/O priority of the whisper process.
• -t, --task <task> Defines the task for whisper to perform. Options are transcribe or translate. Default is transcribe.
• -m, --model <model> Specifies the model for whisper to use. Options include tiny.en, tiny, base.en, base, small.en, small, medium.en, medium, large-v1, large-v2, or large. Default is small.

Note on Whisper Models

Switching between different Whisper models using the -m or --model option necessitates running the whisper command again separately. This step enables Whisper to download the new model before processing files in bulk using the script. For example:

whisper --task transcribe --language en --model medium --output_format txt --output_dir ./ ./myvideo.mp4

This command downloads the medium model if it's not already present on your system. Subsequent uses of the medium model with the script will then proceed without needing to download the model again.

Dependencies

This script's operation relies on Whisper being installed and accessible in your system's PATH. Refer to the official OpenAI Whisper documentation for installation instructions.