douki

Douki is a audio fingerprinting based subtitle synchronization tool.

Douki (同期): synchronization
-- Jisho

This project started in ‎October ‎2020 and was made to automatically synchronize song translations in anime. It was been already tested with dozens of anime, including Ganbare, Douki-chan.

At the moment only a programmatic API for Node is available. This project is experimental and everything can (and probably will) change. Check the roadmap for more info.

Usage

Make sure ffmpeg is installed and available in the path.

Generate synchronization data by running this:

import { generateSyncData } from 'douki'

await generateSyncData({
  // Path to audio or video file containing the section to be synchronized
  sourceFile: 'Test - Episode 2.mkv',
  // Directory where the synchronization data will be stored
  // (a single directory can store data from multiple sections from the same project)
  dataDir: 'projects/test',
  // The start and end times of the section (optional)
  // Can be a seconds number or a 00:00.00 timestamp string
  start: 0,
  end: 90,
  // Name of the section (optional)
  name: 'opening',
  // A .ass file to be used as template (optional)
  templatePath: 'template.ass'
})

By running the above code it will create a .json file containing synchronization data, a .mkv containing the synchronized section, a subdirectory containing attachments found in the source file, an .ass file with the subtitle found in the source file or, if there was not any, based in the template, and a keyframes.txt file.

Edit the .ass as needed. It should be timed against the .mkv file. Run the above code multiple times for each section to be synchronized. The .mkv and keyframes.txt files are meant only to aid editing the .ass file and can be deleted afterwards.

Run the below code to synchronize the section to another files:

import { synchronizeSubtitles } from 'douki'

const result = await synchronizeSubtitles(
  // Path to the new video or audio file to sync subtitles against
  'Test - Episode 3.mkv',
  // Path to the directory where synchronization data was stored
  'projects/test',
  // Output path: can be a directory or an .ass file
  'some/temporary/dir' // or 'some/temporary/file.ass'
)

// The path to the generated subtitle
console.log(result.subtitle)

// An array of paths of the files that need to be
// attached along the generated subtitle
console.log(result.attachments)

// synchronizeSubtitles will return undefined when there are no matches

After that merge the generated subtitles with existent ones and mux with your video.