tytube

A TypeScript port of pytube. Extract YouTube metadata, list streams, and read captions — without the YouTube Data API.

youtube  ·  metadata  ·  streams  ·  captions  ·  playlist  ·  channel  ·  search  ·  no-api-key  ·  pytube  ·  bun  ·  node  ·  esm

Status: pre-release. Metadata extraction, stream discovery, and caption listing all work end-to-end against live YouTube. Full media downloads require a working n-parameter cipher against YouTube's current base.js and are not yet implemented in v0.1 — see Limitations.

Why tytube?

Existing JS YouTube libraries either wrap a Python binary (youtube-dl-exec), depend on the YouTube Data API v3 (rate limits + API key required), or break under non-Node runtimes. tytube is a pure TypeScript rewrite of pytube's API surface that runs natively under Bun, has zero runtime dependencies, and ships ESM + CJS + .d.ts out of the box.

• No Python, no yt-dlp shell-out, no Data API key.
• Zero runtime dependencies — just fetch and (optionally) node:fs.
• 1:1 pytube API parity — if you know pytube, you know tytube.
• First-class Bun support — built and tested on Bun matrix CI.
• Strict TypeScript with full .d.ts declarations and noUncheckedIndexedAccess.

Install

bun add @hallelx/tytube     # Bun
npm install @hallelx/tytube # Node
pnpm add @hallelx/tytube    # pnpm
yarn add @hallelx/tytube    # Yarn

Quick start

import { YouTube } from '@hallelx/tytube';

const yt = new YouTube('https://youtu.be/lpFcNQpH81Q');
await yt.prefetch();

console.log(await yt.title());     // "Build a Language Learning Mobile App: …"
console.log(await yt.author());    // "Andreas Trolle"
console.log(await yt.length());    // 31126 (seconds)
console.log(await yt.views());     // 16372

const streams = await yt.streams();
console.log(`Found ${streams.length} streams`);
for (const stream of streams) {
  console.log(stream.toString());
}

const captions = await yt.captions();
for (const c of captions) {
  console.log(`${c.code}: ${c.name}`);
}

After prefetch(), the *Sync getters work without further I/O:

console.log(yt.titleSync);
console.log(yt.authorSync);
console.log(yt.streamsSync.length);

Playlists, channels, and search

import { Playlist, Channel, Search } from '@hallelx/tytube';

// Playlist iteration — async iterable, lazily paged
const playlist = new Playlist('https://www.youtube.com/playlist?list=PL...');
console.log(await playlist.title());
for await (const video of playlist) {
  console.log(await video.title());
}

// Channel videos
const channel = new Channel('https://www.youtube.com/c/SomeChannel');
console.log(await channel.channelName());
for await (const video of channel) {
  console.log(await video.title());
}

// Search results
const search = new Search('typescript tutorial');
const results = await search.results();
for (const yt of results) {
  console.log(await yt.title());
}

CLI

bunx @hallelx/tytube https://youtu.be/lpFcNQpH81Q --list
bunx @hallelx/tytube https://youtu.be/lpFcNQpH81Q --list-captions
bunx @hallelx/tytube https://youtu.be/lpFcNQpH81Q -c en -t ./captions
bunx @hallelx/tytube --version

Download flags (-r, --itag, -a, -f) are wired up but currently constrained by the cipher limitation below.

API parity with pytube

All metadata getters are async because they may trigger network I/O on first call. After prefetch(), sync *Sync getters work for hot-path access.

| pytube | tytube | | --------------------------------------- | ----------------------------------------------- | | YouTube(url) | new YouTube(url) | | yt.title | await yt.title()  ·  yt.titleSync | | yt.author | await yt.author() | | yt.length | await yt.length() | | yt.views | await yt.views() | | yt.streams | await yt.streams() | | yt.captions | await yt.captions() | | yt.thumbnail_url | await yt.thumbnailUrl() | | yt.channel_url | await yt.channelUrl() | | yt.publish_date | await yt.publishDate() | | yt.metadata | await yt.metadata() | | Playlist(url).videos | for await (const v of new Playlist(url)) | | Channel(url).videos | for await (const v of new Channel(url)) | | Search(q).results | await new Search(q).results() | | stream.download(output_path=…) | await stream.download({ outputPath: '…' }) | | caption.generate_srt_captions() | await caption.generateSrtCaptions() |

Filtering streams works the same way:

const streams = await yt.streams();

streams.getHighestResolution();
streams.getLowestResolution();
streams.getAudioOnly('mp4');
streams.getByItag(22);
streams.getByResolution('720p');
streams.filter({ progressive: true, subtype: 'mp4' }).orderBy('resolution').last();
streams.filter({ onlyAudio: true }).orderBy('abr').last();

Runtime support

| Runtime | Status | | -------------------- | --------------------------------------------------------------------------------------- | | Node.js ≥18 | ✅ Full support | | Bun ≥1.0 | ✅ Full support — built and CI-tested on Bun | | Deno | ✅ Works with --allow-net (uses global fetch) | | Browser | ⚠️ Requires a CORS proxy — YouTube does not send CORS headers | | Cloudflare Workers | ⚠️ Cipher path requires Function() which is blocked under default CSP |

Limitations

Media downloads currently fail with HTTP 403. YouTube ciphers the n parameter on stream URLs, and decoding n requires running JavaScript extracted from YouTube's base.js. tytube's cipher extractor uses pytube's regex patterns, which YouTube's late-2025 base.js has drifted beyond. This is the same problem pytube has hit repeatedly — YouTube rotates base.js and breaks the regex.

The metadata extraction path works completely because it parses ytInitialPlayerResponse directly from the watch HTML, which doesn't depend on cipher decryption.

If you need actual media downloads today, use yt-dlp via child_process.spawn as a temporary bridge while we iterate on the cipher. We're tracking the cipher fix as the headline issue for v0.2.

Contributing

git clone https://github.com/hallelx2/tytube
cd tytube
bun install
bunx vitest run    # 45 tests, all passing
bunx tsup          # build dist/

The package mirrors pytube's module layout 1:1 in src/ so contributors can read both side-by-side. The Python pytube source lives in the parent directory at ../pytube/pytube/ as a read-only reference.

Acknowledgements

• The original pytube project, whose architecture and regex patterns this port closely follows.
• yt-dlp for keeping the InnerTube client configurations current.

License

MIT, with portions derived from pytube (originally The Unlicense).