In short
- captions.list costs 50 quota units and returns a video’s caption tracks without the captions themselves. Read on captions.list (developers.google.com) (opens in a new tab)
- captions.download costs 200 units, needs OAuth with the youtube.force-ssl or youtubepartner scope, and requires the user to have permission to edit the video. Read on captions.download (developers.google.com) (opens in a new tab)
- As of November 2026, a project’s default quota is 100 search.list calls, 100 videos.insert calls and 10,000 units a day for every other method, caption calls included. Read on Quota costs for API requests (developers.google.com) (opens in a new tab)
- PrimaryAPI’s YouTube Video Transcript reads a public video’s existing captions from its URL with one key, and a video without captions costs nothing.
What the YouTube Data API offers for captions
In the Data API, a caption resource is one caption track on a video. Its methods list a video’s tracks, upload a track, update one, download one and delete one. Read on Captions resource, YouTube Data API (developers.google.com) (opens in a new tab)
Each track’s snippet gives its language as a BCP-47 tag, its name, whether it is a draft, and its kind: standard for a regular track, ASR for one made by automatic speech recognition, and forced for one that plays when no other track is selected. Read on Captions resource, YouTube Data API (developers.google.com) (opens in a new tab)
captions.list takes part and videoId, needs OAuth with the youtube.force-ssl or youtubepartner scope, and its docs note that the response does not contain the actual captions. Read on captions.list (developers.google.com) (opens in a new tab)
Why captions.download rarely gets you a transcript
To read a track’s text you call captions.download, and the method requires the user to have permission to edit the video. It works for your own channel or one you manage, not for any public video. Read on captions.download (developers.google.com) (opens in a new tab)
It returns the track in its original format unless you ask for sbv, scc, srt, ttml or vtt, and the tlang parameter returns a machine translation of the track. Read on captions.download (developers.google.com) (opens in a new tab)
For a transcript pipeline that reads videos from many channels, that rules out the official API: you would need edit access to every video you want to read.
How YouTube quota counts caption calls
Google measures Data API use in quota units. As of November 2026, each project gets 100 search.list calls, 100 videos.insert calls and 10,000 units a day for all other methods, and daily quotas reset at midnight Pacific Time. Read on Quota costs for API requests (developers.google.com) (opens in a new tab)
Every request costs at least 1 unit, invalid ones included. Reading a list of resources usually costs 1 unit, and a write usually costs 50. Read on YouTube Data API getting started (developers.google.com) (opens in a new tab)
More quota than the default needs an audit, requested with the YouTube API Services Audit and Quota Extension Form. Read on Quota extension audits (developers.google.com) (opens in a new tab)
| Method | Quota units per call |
|---|---|
| captions.list | 50 |
| captions.insert | 400 |
| captions.update | 450 |
| captions.delete | 50 |
| videos.list | 1 |
| commentThreads.list | 1 |
What a day of quota buys in transcripts
captions.download costs 200 units on top of the 50 for captions.list, so each transcript takes 250 units. A project’s default 10,000 daily units cover 40 transcripts a day, and only for videos the user can edit. Read on captions.download (developers.google.com) (opens in a new tab)
Those units are shared with every video, channel and comment request the project makes, so a busy app has even less left for captions.
Creator captions, automatic captions and no captions
Many videos carry captions made by YouTube’s speech recognition rather than written by the creator. Automatic captions can mishear names and technical terms, so it helps to know which kind you are reading.
PrimaryAPI’s YouTube Video Transcript (Provenance) returns isAutoGenerated with every transcript and accepts captionKind set to manual, automatic or any, along with the video’s title, channel, views and publish time in the same response.
Both YouTube transcript APIs read the captions a video already has and do not run speech-to-text. A video without captions returns no transcript, and because you only pay when data comes back, that call costs nothing.
Getting a transcript with one PrimaryAPI key
Create an account, copy your API key and send a video URL to YouTube Video Transcript. Watch, youtu.be, Shorts, live and embed links all work, or you can pass the video ID.
The JSON holds the video ID, the full transcript as one string, the caption language, the duration and timed segments with start and end seconds. Pass a language code such as en or es to ask for that track.
When links come from several sites, Video Transcript (any platform) also takes TikTok, Instagram, Facebook and X links and returns their transcripts on the same key.
- Summaries: pass the transcript with the title and channel, so the model knows what it is reading.
- Search: index segments with their start times, so a result links to the moment in the video.
- Batch jobs: find videos with YouTube Search, then fetch transcripts for the ones you keep.
- Agents: an agent can find the transcript APIs through the MCP server or SKILL.md and start on a capped trial key that you approve.
When the official captions API is the right tool
If you publish captions on your own channel, the Data API is what you want: it uploads, updates and deletes tracks, and downloads them in the format you choose.
It also fits apps where people sign in with Google and work on their own videos, since they can authorize your app to read and edit their tracks.
Try it with PrimaryAPI
One key for every API below, for your code and your AI agents. You only pay when data comes back.
- YouTube TranscriptExisting captions of a YouTube video or Short as plain text plus timed segments, by URL or ID2 credits / call
- YouTube Video TranscriptTranscript of a YouTube video or Short with timed segments, whether the captions are creator-written or speech recognition, and the video metadata in one response3 credits / call
- Video TranscriptOne public video or post URL in, its transcript out, for YouTube, TikTok, Facebook, Instagram, X / Twitter, LinkedIn; billed only when a transcript is found2 credits / call
- YouTube SearchKeyword search over YouTube videos with relevance or popularity sorting, upload-date filters and cursor pagination2 credits / call
- YouTube VideoMetadata for one YouTube video or Short by URL or ID2 credits / call