Generate-Subtitles

To Install

npm i generate-subtitles

Checkout source code at generate-subtitles repo

Description

This package parses video and audio files, to generate .vtt formatted subtitle files. The package is fully asynchronous and adds support for threads flag to run on lower powered systems complemented by file system support.

The package is built upon

• Introducing Whisper | OpenAI's open source model

• Whisper.cpp | A custom cpp implementation of whisper model in C++.

• whisper-node | A similar package based on OpenAi-whisper and Whisper.cpp

Usage

Note- To run the script for first time the package needs to download the model which is then saved in /node_modules/generate-subtitles/lib/whisper.cpp/models

To do that, run

npx generate-subtitles download

and select an appropriate model.

Usage as modules

• Import generate function from the package.

import { generate } from "generate-subtitles";

• Now make the options data.

const options = {
    //required- where file must be placed
    inputFile: '/path/to/the/file', 
    //optional- when not specified, it saves at the input Dir
    outputDir: '/path/to/the/dir',
    //required- (enum)Valid inputs are only 'audio' and 'video'
    inputType: '/type_of_input(audio/video)',
    //required
    whisperFlags: {
        //not required- currently supports vtt only
        subFormat: 'vtt',
        //required- default model is base
        model: 'model_name',
        //not required- when using a multilingual model
        //default-english
        language: 'language_name',
        //not required- no of threads to run the model
        //default- 7
        threadCount: number,
    }
}

• Pass it to the function, which then outputs the .vtt file to output directory, with same name as input file.

try {
    const getSubtiles= await generate(options);
    //resolves with true
    //An output file is generat
} catch(error) {
   // rejects with an error
}

Miscellaneous

• When using audio input file, it requires .wav format at 16KHz frequency

• In windows based system, make(Make for Windows).

Changelogs

@1.0.6

• Fix: Language had an error, where it took input as model name. @1.0.5

• Fix: removed redundant check input.

• Fix: OutputDir takes a filename, and outputs to that file.

• Fix: removed unecessary lower case checks.

  Acknowledgements

• [Openai-whisper](Introducing Whisper)

• Whisper.cpp

• whisper-node