YouTube · Transcripts

YouTube Video Transcript Scraper API

The YouTube Video Transcript endpoint is one API call: send a URL with your HProxy key and get back a clean, structured JSON object. Pull the full timestamped transcript of any YouTube video, every caption line with its start and end time, in any language. HProxy runs the proxies, sessions, anti-bot handling and parsing server-side, so there is no headless browser to babysit and no parser to keep fixing when YouTube changes its markup. You only ever touch the JSON.

Get your API key YouTube API reference

0.15¢ per call
$1.49 per 1,000
Takes a URL
Errors and timeouts are not charged

On YouTube

Profiles1 Posts and media5Transcripts1Comments2 Search and trends3

Transcripts

YouTube

The call

One GET, and the video transcript comes back as JSON.

Your key goes in one header. 1 required parameter, and the answer is the object below it.

The full reference

curl "https://hproxy.com/api/v1/scrape/youtube/transcript?url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DdQw4w9WgXcQ" \
  -H "X-API-Key: hpx_your_key_here"

Response200 OK · JSON

{
  "language": "en",
  "duration": 47.2,
  "source": "creator",
  "segments": [
    { "start": 0.0, "end": 2.4, "text": "So the first thing we changed was the timing," },
    { "start": 2.4, "end": 5.8, "text": "and that completely changed the result." }
  ]
}

What comes back

The YouTube video transcript object.

segments[]: Each transcript line with its start/end timestamp and text.
language: The detected or declared transcript language.
duration: Total media length the transcript spans.
source: Whether the transcript is creator-provided or machine-generated.

What it takes

Parameters.

urlurlRequiredhttps://www.youtube.com/watch?v=dQw4w9WgXcQ

Target URL (profile, post or video).

languagestringen

Two-letter language code, e.g. en.

cache_max_agestring1d

Accept a cached result up to this old (1d, 3d, 7d, 14d or 30d) and pay nothing for it. Omit for a live scrape. Allowed values: `1d`, `3d`, `7d`, `14d`, `30d`.

How it works

Scraping YouTube video transcript, in three steps.

You write one line. Proxies, anti-bot walls, rendering and parsing happen on our side.

Create a free account Every YouTube endpoint

1
Create a free account
Sign up and top up any amount. The balance pays for every HProxy product.
2
Grab your API key
One key from the dashboard authenticates every scraper.
3
Call the endpoint
Send the GET you need and read the JSON. That is the entire integration.

What people build with it

Where the YouTube video transcript endpoint earns its keep.

Turn YouTube video into searchable, indexable text for content and SEO.
Summarise or translate long YouTube videos at scale.
Build a spoken-content dataset for training or analysis.

YouTube video transcript questions

Answered before you ask.

The reference

Send one authenticated HTTPS request to the YouTube Video Transcript endpoint with a URL and your HProxy API key. Pull the full timestamped transcript of any YouTube video, every caption line with its start and end time, in any language. No proxies, no headless browser and no parsing on your side. The response is ready-to-use JSON.

This endpoint is 0.15¢ per successful call, which is $1.49 per 1,000. It bills from your HProxy balance with no subscription and no monthly minimum. Errors are retried and never charged; a lookup whose target does not exist counts as a call.

The API only accesses publicly available YouTube data, the same pages anyone can open in a browser. You are responsible for using the data in line with applicable laws and YouTube's terms for your use case.

A structured JSON object with the full video transcript data YouTube exposes publicly. See the fields listed above. The schema is consistent across calls, so you integrate once and rely on it.

No. Every call runs on HProxy's own residential network behind the scenes, so requests look like real users. The proxy layer is already included in the per-call price.

Docs and AI

Connect your scraper to an AI, any time.

The whole Scraper API is published as markdown, so an assistant reads every endpoint, every parameter and a real response without a key and writes your integration. And because a scrape is one GET that answers in JSON, an agent can run the scraper itself, on its own schedule, with your key.

Read the scraper docs Set up your assistant

Claude, ChatGPT, Cursor, Gemini, Copilot, Windsurf, Perplexity, Ollama and any assistant that speaks MCP.

The API as markdown

Paste /llms/scraper-api.txt into Claude, ChatGPT, Cursor or Copilot: every endpoint and the whole catalog, in one file. No key and no account to read it.

Open the markdown

An agent can call it

One GET with your key returns structured JSON, so an agent scrapes on its own schedule. No browser to drive, no parser to keep alive between runs.

See the endpoints

MCP server and CLI

Built by us for the free tools and hosted at mcp.hproxy.com: the proxy list, the checker and the IP lookup as tools in Claude, ChatGPT, Cursor or any assistant that speaks MCP. Nothing to install, no key.

Add it to your assistant