Async YouTube Transcript (youscript)

An asyncio-based YouTube transcript fetcher that retrieves multiple videos concurrently, exposing both a Python API and a small CLI.

Features

• Concurrent transcript fetching for many videos
• Manual and auto-generated caption support
• Simple Python API with async context manager
• Minimal CLI for quick one-off usage

Requirements

• Python 3.12+
• uv (for environment and workflows)

Installation (uv)

• Create/sync the project environment:

  • uv sync

• Ensure the package is importable (editable install) if needed in your workflow:

  • uv pip install -e .

Quick Start

CLI

• Fetch English transcripts for a single video and save as text:
  • uv run ust "https://www.youtube.com/watch?v=VIDEO_ID" -l en -s txt

Options

• -l/--languages: repeatable, default en
• -s/--save-as: output format, default txt
• -p/--proxy: proxy URL (optional)

Python API

import asyncio
from youscript.client import YouScript

urls = [
    "https://www.youtube.com/watch?v=HyzlYwjoXOQ",
    "https://www.youtube.com/watch?v=0VLAoVGf_74",
]

async def main():
    async with YouScript() as client:
        results = await client.get_transcripts(urls, languages=["en"])
        for url, transcripts in results.items():
            print(url, "->", len(transcripts), "transcript variants")

asyncio.run(main())

Benchmarking

• Summary: the async client is substantially faster than the non-async youtube-transcript-api for batch fetching.
• Why: transcript retrieval is network-bound; async overlaps I/O so total time approaches the slowest single request, not the sum of all.

Reproduce locally with uv

• Install deps and run the benchmark in one shot:
  • uv run --with youtube-transcript-api scripts/benchmark.py
• The script prints two lines:
  • Async time elapsed: <seconds>
  • Sequential time elapsed: <seconds>

Snapshot (example run)

Notes

• Sequential baseline: youtube-transcript-api .fetch(...) called one-by-one.
• Async path: our client fetches concurrently using asyncio.TaskGroup.
• Exact speedup varies with network/YouTube latency; expect notable gains on batches.