InnerTube is an API used by all YouTube clients. It was created to simplify the deployment of new features and experiments across the platform ^1. This library manages all low-level communication with InnerTube, providing a simple and efficient way to interact with YouTube programmatically. Its design aims to closely emulate an actual client, including the parsing of API responses.

If you have any questions or need help, feel free to reach out to us on our Discord server or open an issue here.

Table of Contents

1. Prerequisites
2. Installation
3. Usage
   • Browser Usage
   • Caching
   • API
   • Extending the library
4. Contributing
5. Contact
6. Disclaimer
7. License

Prerequisites

YouTube.js runs on Node.js, Deno, and modern browsers.

It requires a runtime with the following features:

• fetch
  • On Node, we use undici’s fetch implementation, which requires Node.js 16.8+. If you need to use an older version, you may provide your own fetch implementation. See providing your own fetch implementation for more information.
  • The Response object returned by fetch must thus be spec compliant and return a ReadableStream object if you want to use the VideoInfo#download method. (Implementations like node-fetch return a non-standard Readable object.)
• EventTarget and CustomEvent are required.

Installation

# NPM
npm install youtubei.js@latest

# Yarn
yarn add youtubei.js@latest

# Git (edge version)
npm install github:LuanRT/YouTube.js

When using Deno, you can import YouTube.js directly from deno.land:

import { Innertube } from 'https://deno.land/x/youtubei/deno.ts';

Usage

Create an InnerTube instance:

// const { Innertube } = require('youtubei.js');
import { Innertube } from 'youtubei.js';
const youtube = await Innertube.create(/* options */);

Options

Browser Usage

To use YouTube.js in the browser, you must proxy requests through your own server. You can see our simple reference implementation in Deno at examples/browser/proxy/deno.ts.

You may provide your own fetch implementation to be used by YouTube.js, which we will use to modify and send the requests through a proxy. See examples/browser/web for a simple example using Vite.

// Multiple exports are available for the web.
// Unbundled ESM version
import { Innertube } from 'youtubei.js/web';
// Bundled ESM version
// import { Innertube } from 'youtubei.js/web.bundle';
// Production Bundled ESM version
// import { Innertube } from 'youtubei.js/web.bundle.min';
await Innertube.create({
  fetch: async (input: RequestInfo | URL, init?: RequestInit) => {
    // Modify the request
    // and send it to the proxy

    // fetch the URL
    return fetch(request, init);
  }
});

Streaming

YouTube.js supports streaming of videos in the browser by converting YouTube’s streaming data into an MPEG-DASH manifest.

The example below uses dash.js to play the video.

import { Innertube } from 'youtubei.js/web';
import dashjs from 'dashjs';

const youtube = await Innertube.create({ /* setup - see above */ });

// Get the video info
const videoInfo = await youtube.getInfo('videoId');

// now convert to a dash manifest
// again - to be able to stream the video in the browser - we must proxy the requests through our own server
// to do this, we provide a method to transform the URLs before writing them to the manifest
const manifest = await videoInfo.toDash(url => {
  // modify the url
  // and return it
  return url;
});

const uri = "data:application/dash+xml;charset=utf-8;base64," + btoa(manifest);

const videoElement = document.getElementById('video_player');

const player = dashjs.MediaPlayer().create();
player.initialize(videoElement, uri, true);

A fully working example can be found in examples/browser/web.

Providing your own fetch implementation

You may provide your own fetch implementation to be used by YouTube.js. This can be useful in some cases to modify the requests before they are sent and transform the responses before they are returned (eg. for proxies).

// provide a fetch implementation
const yt = await Innertube.create({
  fetch: async (input: RequestInfo | URL, init?: RequestInit) => {
    // make the request with your own fetch implementation
    // and return the response
    return new Response(
      /* ... */
    );
  }
});

Caching

Caching the transformed player instance can greatly improve the performance. Our UniversalCache implementation uses different caching methods depending on the environment.

In Node.js, we use the node:fs module, Deno.writeFile() in Deno, and indexedDB in browsers.

By default, the cache stores data in the operating system’s temporary directory (or indexedDB in browsers). You can make this cache persistent by specifying the path to the cache directory, which will be created if it doesn’t exist.

import { Innertube, UniversalCache } from 'youtubei.js';
// Create a cache that stores files in the OS temp directory (or indexedDB in browsers) by default.
const yt = await Innertube.create({
  cache: new UniversalCache(false)
});

// You may want to create a persistent cache instead (on Node and Deno).
const yt = await Innertube.create({
  cache: new UniversalCache(
    // Enables persistent caching
    true, 
    // Path to the cache directory. The directory will be created if it doesn't exist
    './.cache' 
  )
});

API

• Innertube

  Properties

  • .session
  • .account
  • .interact
  • .playlist
  • .music
  • .studio
  • .kids

  Methods

  • .getInfo(target, client?)
  • .getBasicInfo(video_id, client?)
  • .search(query, filters?)
  • .getSearchSuggestions(query)
  • .getComments(video_id, sort_by?)
  • .getHomeFeed()
  • .getGuide()
  • .getLibrary()
  • .getHistory()
  • .getTrending()
  • .getSubscriptionsFeed()
  • .getChannel(id)
  • .getNotifications()
  • .getUnseenNotificationsCount()
  • .getPlaylist(id)
  • .getHashtag(hashtag)
  • .getStreamingData(video_id, options)
  • .download(video_id, options?)
  • .resolveURL(url)
  • .call(endpoint, args?)

getInfo(target, client?)

Retrieves video info.

Returns: Promise<VideoInfo>

getBasicInfo(video_id, client?)

Suitable for cases where you only need basic video metadata. Also, it is faster than getInfo().

Returns: Promise<VideoInfo>

search(query, filters?)

Searches the given query on YouTube.

Returns: Promise<Search>

Note Search extends the Feed class.

getSearchSuggestions(query)

Retrieves search suggestions for given query.

Returns: Promise<string[]>

getComments(video_id, sort_by?)

Retrieves comments for given video.

Returns: Promise<Comments>

See ./examples/comments for examples.

getHomeFeed()

Retrieves YouTube’s home feed.

Returns: Promise<HomeFeed>

Note HomeFeed extends the FilterableFeed class.

getGuide()

Retrieves YouTube’s content guide.

Returns: Promise<Guide>

getLibrary()

Retrieves the account’s library.

Returns: Promise<Library>

Note Library extends the Feed class.

getHistory()

Retrieves watch history.

Returns: Promise<History>

Note History extends the Feed class.

getTrending()

Retrieves trending content.

Returns: Promise<TabbedFeed<IBrowseResponse>>

getSubscriptionsFeed()

Retrieves the subscriptions feed.

Returns: Promise<Feed<IBrowseResponse>>

getChannel(id)

Retrieves contents for a given channel.

Returns: Promise<Channel>

Note Channel extends the TabbedFeed class.

See ./examples/channel for examples.

getNotifications()

Retrieves notifications.

Returns: Promise<NotificationsMenu>

getUnseenNotificationsCount()

Retrieves unseen notifications count.

Returns: Promise<number>

getPlaylist(id)

Retrieves playlist contents.

Returns: Promise<Playlist>

Note Playlist extends the Feed class.

getHashtag(hashtag)

Retrieves a given hashtag’s page.

Returns: Promise<HashtagFeed>

Note HashtagFeed extends the FilterableFeed class.

getStreamingData(video_id, options)

Returns deciphered streaming data.

Note This method will be deprecated in the future. We recommend retrieving streaming data from a VideoInfo or TrackInfo object instead if you want to select formats manually. Please refer to the following example:

const info = await yt.getBasicInfo('somevideoid');

const url = info.streaming_data?.formats[0].decipher(yt.session.player);
console.info('Playback url:', url);

// or:
const format = info.chooseFormat({ type: 'audio', quality: 'best' });
const url = format?.decipher(yt.session.player);
console.info('Playback url:', url);

Returns: Promise<object>

download(video_id, options?)

Downloads a given video.

Returns: Promise<ReadableStream<Uint8Array>>

See ./examples/download for examples.

resolveURL(url)

Resolves a given url.

Returns: Promise<NavigationEndpoint>

call(endpoint, args?)

Utility to call navigation endpoints.

Returns: Promise<T extends IParsedResponse | IParsedResponse | ApiResponse>

Extending the library

YouTube.js is modular and easy to extend. Most of the methods, classes, and utilities used internally are exposed and can be used to implement your own extensions without having to modify the library’s source code.

For example, let’s say we want to implement a method to retrieve video info. We can do that by using an instance of the Actions class:

import { Innertube } from 'youtubei.js';

(async () => {
  const yt = await Innertube.create();

  async function getVideoInfo(videoId: string) {
    const videoInfo = await yt.actions.execute('/player', {
      // You can add any additional payloads here, and they'll merge with the default payload sent to InnerTube.
      videoId,
      client: 'YTMUSIC', // InnerTube client to use.
      parse: true // tells YouTube.js to parse the response (not sent to InnerTube).
    });

    return videoInfo;
  }

  const videoInfo = await getVideoInfo('jLTOuvBTLxA');
  console.info(videoInfo);
})();

Alternatively, suppose we locate a NavigationEndpoint in a parsed response and want to see what happens when we call it:

import { Innertube, YTNodes } from 'youtubei.js';

(async () => {
  const yt = await Innertube.create();

  const artist = await yt.music.getArtist('UC52ZqHVQz5OoGhvbWiRal6g');
  const albums = artist.sections[1].as(YTNodes.MusicCarouselShelf);

  // Let's imagine that we wish to click on the “More” button:
  const button = albums.as(YTNodes.MusicCarouselShelf).header?.more_content;

  if (button) {
    // Having ensured that it exists, we can then call its navigation endpoint using the following code:
    const page = await button.endpoint.call(yt.actions, { parse: true });
    console.info(page);
  }
})();

Parser

YouTube.js’ parser enables you to parse InnerTube responses and convert their nodes into strongly-typed objects that are simple to manipulate. Additionally, it provides numerous utility methods that make working with InnerTube a breeze.

Here’s an example of its usage:

// See ./examples/parser

import { Parser, YTNodes } from 'youtubei.js';
import { readFileSync } from 'fs';

// YouTube Music's artist page response
const data = readFileSync('./artist.json').toString();

const page = Parser.parseResponse(JSON.parse(data));

const header = page.header?.item().as(YTNodes.MusicImmersiveHeader, YTNodes.MusicVisualHeader);

console.info('Header:', header);

// The parser uses a proxy object to add type safety and utility methods for working with InnerTube's data arrays:
const tab = page.contents?.item().as(YTNodes.SingleColumnBrowseResults).tabs.firstOfType(YTNodes.Tab);

if (!tab)
  throw new Error('Target tab not found');

if (!tab.content)
  throw new Error('Target tab appears to be empty');
  
const sections = tab.content?.as(YTNodes.SectionList).contents.as(YTNodes.MusicCarouselShelf, YTNodes.MusicDescriptionShelf, YTNodes.MusicShelf);

console.info('Sections:', sections);

Documentation for the parser can be found here.

Contributing

We welcome all contributions, issues and feature requests, whether small or large. If you want to contribute, feel free to check out our issues page and our guidelines.

We are immensely grateful to all the wonderful people who have contributed to this project. A special shoutout to all our contributors! 🎉

Contact

LuanRT - @thesciencephile - luanrt@thatsciencephile.com

Project Link: https://github.com/LuanRT/YouTube.js

Disclaimer

This project is not affiliated with, endorsed, or sponsored by YouTube or any of its affiliates or subsidiaries. All trademarks, logos, and brand names used in this project are the property of their respective owners and are used solely to describe the services provided.

As such, any usage of trademarks to refer to such services is considered nominative use. If you have any questions or concerns, please contact me directly via email.

License

Distributed under the MIT License.

(back to top)