Video transcript

When to use this

Use it to turn a video into text for summaries, search, research or repurposing. It reads the video's captions: the creator's own or YouTube's auto-generated ones (isGenerated tells them apart).

Pick a language with language (e.g. es); without it you get English when the video has it, else the video's main language. availableLanguages lists every caption track. YouTube's machine translation isn't available, so a language the video has no captions in returns 404.

Videos without captions, and private or removed ones, return 404 not_found.

Example request

Copy one and replace YOUR_API_KEY with your API key (get one free). The code versions read the key from the LURKAPI_API_KEY environment variable instead.

curl "https://api.lurkapi.com/v1/youtube/video/transcript?url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DdQw4w9WgXcQ" \
  -H "x-api-key: YOUR_API_KEY"

const params = new URLSearchParams({
  url: "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
});
const res = await fetch(`https://api.lurkapi.com/v1/youtube/video/transcript?${params}`, {
  headers: { "x-api-key": process.env.LURKAPI_API_KEY },
});
const data = await res.json();
if (!data.success) throw new Error(`${data.code}: ${data.error}`);
console.log(data.transcript);

import os
import requests

res = requests.get(
    "https://api.lurkapi.com/v1/youtube/video/transcript",
    params={
        "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    },
    headers={"x-api-key": os.environ["LURKAPI_API_KEY"]},
    timeout=60,
)
data = res.json()
if not data["success"]:
    raise RuntimeError(f"{data['code']}: {data['error']}")
print(data["transcript"])

package main

import (
	"encoding/json"
	"fmt"
	"net/http"
	"net/url"
	"os"
	"time"
)

func main() {
	params := url.Values{}
	params.Set("url", "https://www.youtube.com/watch?v=dQw4w9WgXcQ")
	req, _ := http.NewRequest("GET", "https://api.lurkapi.com/v1/youtube/video/transcript?"+params.Encode(), nil)
	req.Header.Set("x-api-key", os.Getenv("LURKAPI_API_KEY"))

	client := &http.Client{Timeout: 60 * time.Second}
	res, err := client.Do(req)
	if err != nil {
		panic(err)
	}
	defer res.Body.Close()

	var data map[string]any
	if err := json.NewDecoder(res.Body).Decode(&data); err != nil {
		panic(err)
	}
	if data["success"] != true {
		panic(fmt.Sprintf("%v: %v", data["code"], data["error"]))
	}
	out, _ := json.MarshalIndent(data["transcript"], "", "  ")
	fmt.Println(string(out))
}

<?php
$params = http_build_query([
    'url' => 'https://www.youtube.com/watch?v=dQw4w9WgXcQ',
]);
$context = stream_context_create(['http' => [
    'header' => 'x-api-key: ' . getenv('LURKAPI_API_KEY'),
    'timeout' => 60,
    'ignore_errors' => true,
]]);
$body = file_get_contents('https://api.lurkapi.com/v1/youtube/video/transcript?' . $params, false, $context);
$data = json_decode($body, true);
if (!$data['success']) {
    throw new Exception("{$data['code']}: {$data['error']}");
}
print_r($data['transcript']);

require 'json'
require 'net/http'

uri = URI('https://api.lurkapi.com/v1/youtube/video/transcript')
uri.query = URI.encode_www_form(
  'url' => 'https://www.youtube.com/watch?v=dQw4w9WgXcQ',
)
req = Net::HTTP::Get.new(uri)
req['x-api-key'] = ENV.fetch('LURKAPI_API_KEY')
res = Net::HTTP.start(uri.host, uri.port, use_ssl: uri.scheme == 'https', read_timeout: 60) do |http|
  http.request(req)
end
data = JSON.parse(res.body)
raise "#{data['code']}: #{data['error']}" unless data['success']
puts JSON.pretty_generate(data['transcript'])

Parameters

Add them to the URL after a ?. Join several with &.

Example response

A real response, captured from the live API. Lists are trimmed to a couple of items, and long text such as media links is cut short with ….

{
  "success": true,
  "credits_remaining": 999983,
  "credits_charged": 1,
  "videoId": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "language": "English",
  "languageCode": "en",
  "isGenerated": false,
  "transcript": [
    {
      "text": "♪ We're no strangers to love ♪",
      "startMs": 18640,
      "endMs": 21880,
      "startTimeText": "0:18"
    },
    {
      "text": "♪ You know the rules and so do I ♪",
      "startMs": 22640,
      "endMs": 26960,
      "startTimeText": "0:22"
    }
  ],
  "transcript_only_text": "♪ We're no strangers to love ♪ ♪ You know the rules and so do I ♪",
  "availableLanguages": [
    {
      "code": "en",
      "name": "English",
      "isGenerated": false
    },
    {
      "code": "en",
      "name": "English",
      "isGenerated": true
    },
    {
      "code": "de-DE",
      "name": "German (Germany)",
      "isGenerated": false
    },
    {
      "code": "ja",
      "name": "Japanese",
      "isGenerated": false
    },
    {
      "code": "pt-BR",
      "name": "Portuguese (Brazil)",
      "isGenerated": false
    },
    {
      "code": "es-419",
      "name": "Spanish (Latin America)",
      "isGenerated": false
    }
  ]
}

Response fields

Every field in a successful response, and what it means. A name with [] belongs to each item of a list: transcript[].text is the text of every item in transcript.

Cost and freshness

Each call costs 1 credit. Calls turned away before they reach the platform (invalid parameters, a key or balance problem, a rate limit) are free, and so are 5xx errors. A not_found is charged, since the platform was checked. How charging works.

Answers are cached for up to 30 days. A repeat call within that time can return the same data, and costs the same.

Errors

A failed call returns success: false, an error message, a code your program can check, and a docs link. These are the errors this endpoint can return. All error codes.

Use it in Claude

Connect LurkAPI to Claude and ask in plain English. This endpoint is the youtube_transcript tool. For example:

Use LurkAPI's youtube_transcript tool with url "https://www.youtube.com/watch?v=dQw4w9WgXcQ", then summarize what you find.

Claude calls youtube_transcript with arguments like these, and each call costs 1 credit:

{"url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ"}

To save tokens, tool results leave out empty fields, keep one URL per image or video, and give transcripts as plain text unless you pass timestamps=true. A result over 45,000 characters comes in parts. Its _note gives the offset for the next part, and later parts are free for 15 minutes. Pass trim=false for the full response.