Blog/Can the YouTube Data API download videos? No, and here is what it does
OverviewAll posts
Guide

Can the YouTube Data API download videos? No, and here is what it does

The YouTube Data API returns metadata, never the media file. See what each official endpoint gives you and how to obtain the file for a pipeline.

VTornado API team
Covered in this article
What the Data API returns
Why there is no download endpoint
How to obtain the file in a pipeline
5 min reading time
Published November 14, 2026
VProduct guides by Velys Software

No. The YouTube Data API v3 has no endpoint that returns a video or audio file. It describes videos, channels and playlists as JSON. If your application needs the media itself, for transcription, editing or analysis, the official API gives you the identifier and the metadata, and the file has to come from somewhere else.

What the Data API returns

EndpointWhat you getMedia file
videos.listTitle, description, duration, statistics, status, available caption track listNo
search.listVideo, channel and playlist IDs matching a queryNo
playlistItems.listThe video IDs of a playlistNo
channels.listChannel details and its uploads playlist IDNo
captions.downloadA caption track, only for videos you own or manage, with OAuthNo

Two points cause most of the confusion:

  • contentDetails describes the file, with fields such as duration, definition and dimension. It describes; it does not link to the bytes.
  • fileDetails and processingDetails exist, and they are only returned to the owner of the video. They report on the uploaded file and still contain no download URL.

The embedded player and the IFrame API play a video in a page. They do not expose a file either.

Why there is no download endpoint

The API is built for managing and displaying YouTube content: search, playlists, channel administration, embeds. Its terms of service do not allow downloading or storing audiovisual content through it. The quota system reflects the same design: quota is counted per metadata request, and nothing in it measures bytes of media.

For your own videos, the supported way to retrieve originals is YouTube Studio or Google Takeout, which are manual and not an API for a product.

A sound division of work

The Data API remains the right tool for everything it does well. A typical pipeline uses both sides:

  1. Discover and filter with the Data API. Resolve a channel to its uploads playlist, page through playlistItems.list, and keep the video IDs you need. Use videos.list to drop videos that are too long, still live or not public.
  2. Obtain the files with a download service or your own tooling. This is the step the Data API does not cover.
  3. Process the files downstream. Transcription, clipping, indexing and training stay in your application.

Filtering in step 1 saves money in step 2: a video you exclude by duration or status is a file you never download.

Obtaining the file

You can operate an open source downloader yourself. It works well on one machine and becomes operational work on a server, as described in yt-dlp not working: a ten-minute triage.

Or you can submit the video URL to a job API that returns the file. With Tornado API:

curl -X POST "https://api.tornadoapi.io/jobs" \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.youtube.com/watch?v=VIDEO_ID", "format": "mp4"}'

The response contains a job_id. Poll GET /jobs/{id} or pass a webhook_url, and collect the file from your storage when the status is Completed. One request can also create up to 100 jobs, which fits the list of IDs produced by step 1. The Python tutorial shows submission, polling and resuming after a restart.

Eligible new accounts get 100 GB free for 7 days, and failed jobs are not billed.

Rights and limits

A technical ability to obtain a file does not grant the right to use it. Process content you own, have licensed or are otherwise authorized to use, and check the terms that apply to your case. Private, removed and restricted videos stay unavailable whichever method you choose.

Sources