YouTube Transcripts API
Scrape YouTube Transcripts data with one API call. Batch transcript lookup for LLM corpora. POST a JSON body with an `ids` array of 1–100 BARE 11-character YouTube video ids (full watch URLs are not accepted here — use GET /v1/youtube/video/transcript for a single URL); returns one row per id — the transcript (`format: "text"` joins it into one string; `format: "segments"` returns `[{start_ms, duration_ms, text}]`), the resolved `language`, and a `status` of ok / not_found / error / deferred — in input order. Optional `language` picks a preferred caption track. Per-id isolation: one caption-less or deleted video never fails the batch — a caption-less video is `not_found` with `ext.reason: "no_captions"`, a deleted/unavailable one is `not_found` with `ext.reason: "video_gone"`. Billing is per successful row at 3 credits (the single-transcript registry tier); not_found / errored / deferred rows are refunded, so you pay exactly for the transcripts you got. Never cached. Optionally streams Server-Sent Events when the client sends `Accept: text/event-stream`.
Last updated August 2026Maintained by the SocialCrawl team
Returns one row per video id, up to 100 per request, each with the spoken transcript, the caption language it resolved to, and an ok or not-found status.
Use it to build a transcript set from many videos at once; a video without captions comes back as not found instead of failing the whole request.
Searching 48 platforms in parallel
TikTok·
Instagram·
YouTube·
Facebook·
X·
LinkedIn·
Reddit·
Threads·
Pinterest·
Twitch·
Truth Social·
Snapchat·
Kick·
Bluesky·
Kwai·
Rumble·
Spotify·TikTok Shop·
Amazon Shop·
Google Shopping·
Trustpilot·
TripAdvisor·
Linktree·
Komi·
Pillar·
lnk.bio·Facebook Ads·Google Ads·LinkedIn Ads·Google Search·Google News·Google Finance·
Polymarket·
Tavily·
Hacker News·
GitHub·
Perplexity·
Naver·Utility·
Universal SearchTikTok·Instagram·YouTube·Facebook·X·LinkedIn·Reddit·Threads·Pinterest·Twitch·Truth Social·Snapchat·Kick·Bluesky·Kwai·Rumble·Spotify·TikTok Shop·Amazon Shop·Google Shopping·Trustpilot·TripAdvisor·Linktree·Komi·Pillar·lnk.bio·Facebook Ads·Google Ads·LinkedIn Ads·Google Search·Google News·Google Finance·Polymarket·Tavily·Hacker News·GitHub·Perplexity·Naver·Utility·Universal SearchTikTok·Instagram·YouTube·Facebook·X·LinkedIn·Reddit·Threads·Pinterest·Twitch·Truth Social·Snapchat·Kick·Bluesky·Kwai·Rumble·Spotify·TikTok Shop·Amazon Shop·Google Shopping·Trustpilot·TripAdvisor·Linktree·Komi·Pillar·lnk.bio·Facebook Ads·Google Ads·LinkedIn Ads·Google Search·Google News·Google Finance·Polymarket·Tavily·Hacker News·GitHub·Perplexity·Naver·Utility·Universal SearchTikTok·Instagram·YouTube·Facebook·X·LinkedIn·Reddit·Threads·Pinterest·Twitch·Truth Social·Snapchat·Kick·Bluesky·Kwai·Rumble·Spotify·TikTok Shop·Amazon Shop·Google Shopping·Trustpilot·TripAdvisor·Linktree·Komi·Pillar·lnk.bio·Facebook Ads·Google Ads·LinkedIn Ads·Google Search·Google News·Google Finance·Polymarket·Tavily·Hacker News·GitHub·Perplexity·Naver·Utility·Universal SearchWhat can you do with the Transcripts API?
The Transcripts endpoint gives you structured YouTube data with computed fields in a single request. No scraping infrastructure to build or maintain.
Example Request
curl -X POST -H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"ids":"dQw4w9WgXcQ,9bZkp7q19f0"}' \
"https://www.socialcrawl.dev/v1/youtube/transcripts"import requests
response = requests.post(
"https://www.socialcrawl.dev/v1/youtube/transcripts",
json={
'ids': 'dQw4w9WgXcQ,9bZkp7q19f0',
},
headers={"x-api-key": "YOUR_API_KEY"},
)
data = response.json()const response = await fetch(
"https://www.socialcrawl.dev/v1/youtube/transcripts",
{
method: "POST",
headers: {
"x-api-key": "YOUR_API_KEY",
"Content-Type": "application/json",
},
body: JSON.stringify({
ids: "dQw4w9WgXcQ,9bZkp7q19f0",
}),
},
);
const data = await response.json();Parameters
| Parameter | Required | Description |
|---|---|---|
| ids | Yes | JSON array of 1–100 BARE YouTube video ids (11 characters each, e.g. 'dQw4w9WgXcQ'). Full watch URLs are not accepted here. |
| language | No | Preferred caption-track language (2-letter code, e.g. 'en', 'ko'). Falls back to the default track when absent or unavailable. |
| format | No | `text` (default) returns the whole transcript as one joined string; `segments` returns timestamped `[{start_ms, duration_ms, text}]`. |
What does the YouTube Transcripts API return?
Every response follows one unified schema. Here is a real, unmodified response body, so you can see the exact fields you get back before spending a credit.
Example response
{
"success": true,
"platform": "tiktok",
"endpoint": "/v1/tiktok/post/transcript",
"data": {
"id": "7499229683859426602",
"url": "https://www.tiktok.com/@stoolpresidente/video/7499229683859426602",
"transcript": "WEBVTT\n\n\n00:00:00.180 --> 00:00:01.740\nAlright, pizza review time.\n\n00:00:01.741 --> 00:00:03.701\nSal's Pizza Factory. Oh,\n\n00:00:03.702 --> 00:00:05.701\nit's fucking pouring. We're in Charlotte.\n\n00:00:05.702 --> 00:00:07.781\nAlright man, any chance I get a real quick picture?\n\n00:00:07.782 --> 00:00:09.901\nJack or Dwight Howard? I gotta pick one.\n\n00:00:09.902 --> 00:00:10.861\nYeah, Jack.\n\n00:00:10.862 --> 00:00:12.301\nYeah, easy, right?\n\n00:00:12.302 --> 00:00:13.981\nWe're still Jason Williams still with us?\n\n00:00:13.982 --> 00:00:16.381\nYeah, he on the plane right over,\n\n00:00:16.382 --> 00:00:18.261\nout of the blue. Said the following.\n\n00:00:18.262 --> 00:00:21.141\nHe would take Dwight Howard over Shaq at center.\n\n00:00:21.142 --> 00:00:23.461\nAnd says he lie about it for the rest of his life\n\n00:00:23.462 --> 00:00:25.341\nand would never admit to saying that.\n\n00:00:25.342 --> 00 …"
},
"credits_used": 10,
"credits_remaining": 9999,
"request_id": "req-8Kq2ZmR4vT9xLb3P",
"cached": false
}Example captured from the TikTok API. Every SocialCrawl endpoint returns this same unified schema, so your YouTube Transcripts response has the same fields.
How does the YouTube Transcripts API work?
Send a GET request with your API key and get back clean, structured JSON in our unified schema. Supported computed fields are populated when the source provides the required inputs.
Method
POST
Response
JSON
How do you scrape social media data in seconds?
The fastest social media scraping API for developers. Scrape profiles, posts, comments, and analytics from 48 platforms covering 10B+ monthly active users.
One schema, every platform
Query 48 platforms with identical response structures. Write your integration once.
Computed fields, not just scraped
When an endpoint supports these metrics and the source provides the required inputs, the normalized record includes engagement_rate, estimated_reach, content_category, and language — ready to use.
See your data before you code
Visual Data Explorer — paste any URL, get rich result cards, sortable tables, CSV export.
import requests
response = requests.get(
'https://www.socialcrawl.dev/v1/tiktok/profile',
params={'handle': 'charlidamelio'},
headers={'x-api-key': 'sc_YOUR_API_KEY'}
)
data = response.json(){
"success": true,
"platform": "tiktok",
"data": {
"author": {
"username": "charlidamelio",
"followers": 152400000
},
"engagement": {
"likes": 12400000000,
"engagement_rate": 0.087
},
"metadata": {
"language": "en",
"content_category": "lifestyle"
}
}
}Ready to scrape YouTube Transcripts data?
Get your API key and start pulling YouTube data in under 60 seconds.