YouTube Transcript Scraper

Get the spoken transcript of any YouTube video — by URL, by channel, or by searching a keyword. No API key, no login, no cookies.

$2.99 per 1,000 transcripts. You are only charged for transcripts actually returned.

---

Why this one

Most transcript scrapers ask YouTube for English and give up when there isn't one. A Korean beauty video with a Korean transcript comes back empty, and you are left guessing whether the video had captions at all.

This Actor does two things differently.

It falls back instead of failing. If your preferred language is not available, you get whatever transcript the video does have, preferring a human-written one over an auto-generated one. Across 39 long-form videos in a mixed English/Korean sample, 39 returned a transcript.

It tells you why when there is nothing. Every row carries transcript_status. An empty transcript is never ambiguous:

transcript_status	What it means
ok	Transcript returned
disabled_by_uploader	The uploader switched captions off. Nobody can retrieve this one.
no_transcript	The video has no caption track at all
video_unavailable	Private, deleted, or age-restricted
blocked	YouTube rate-limited the request
error	Something else went wrong

---

Three ways in

1. Specific videos

{
  "videoUrls": [
    "https://www.youtube.com/watch?v=OrElyY7MFVs",
    "https://youtu.be/dQw4w9WgXcQ",
    "https://www.youtube.com/shorts/58mOSOMn72g"
  ]
}

Watch links, Shorts links, youtu.be links and bare 11-character IDs all work.

2. A whole channel

{
  "channelUrls": ["@aliabdaal", "https://www.youtube.com/@veritasium"],
  "maxVideosPerSource": 25
}

3. Everything on a topic

{
  "keywords": ["korean skincare", "glass skin"],
  "maxVideosPerSource": 20
}

Searches YouTube and transcribes the results. Useful when you want the conversation on a subject rather than one creator.

---

Worked examples

Feed an LLM or RAG index — full text only, no timestamps

{
  "channelUrls": ["@lexfridman"],
  "maxVideosPerSource": 50,
  "includeSegments": false
}

Dropping segments makes the dataset far smaller when you only need the words.

Build subtitles — keep the timings

{
  "videoUrls": ["https://www.youtube.com/watch?v=OrElyY7MFVs"],
  "includeSegments": true
}

Each line comes back as {"start": 0.24, "duration": 3.84, "text": "..."}.

Non-English content

{
  "keywords": ["메이크업"],
  "languages": ["ko", "en"],
  "maxVideosPerSource": 30
}

Long-form only, highest hit rate

{
  "keywords": ["web scraping tutorial"],
  "skipShorts": true,
  "maxVideosPerSource": 30
}

Most Shorts have captions switched off by their uploader. Skipping them raises the share of videos that come back with text.

Research a competitor's whole channel

{
  "channelUrls": ["@competitor"],
  "maxVideosPerSource": 200,
  "includeSegments": false
}

Python

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])

run = client.actor("reportable_broth/youtube-transcript-scraper").call(run_input={
    "keywords": ["korean skincare"],
    "languages": ["ko", "en"],
    "maxVideosPerSource": 20,
})

for r in client.dataset(run.default_dataset_id).iterate_items():
    if r["transcript_status"] == "ok":
        print(f'{r["transcript_language"]:>6}  {len(r["transcript"]):>6} chars  {r["title"]}')
    else:
        print(f'  skipped ({r["transcript_status"]}): {r["title"]}')

On apify-client 3.x the object returned by .call() is a model, not a dict. Use run.default_dataset_id, not run["defaultDatasetId"].

---

Output

Field	Type	Notes
video_id	string
url	string	Watch link
title	string
channel	string or null	Channel name
channel_id	string or null
channel_username	string or null	The @handle
channel_thumbnail	string or null	Channel avatar
subscriber_count	integer or null	Read from the video owner block; YouTube omits it on some page variants
subscriber_count_text	string or null	As YouTube shows it, e.g. 6.69M subscribers
thumbnail	string or null	Video thumbnail
published_at	string or null	Exact ISO 8601 publish time
published_ts	integer or null	The same, as a Unix timestamp
duration_seconds	integer or null
view_count	integer or null	Exact
like_count	integer or null	Exact
comment_count_text	string or null	As YouTube shows it, e.g. 4.2K
comment_count_approx	integer or null	That text as a number — rounded
comment_count	integer or null	Exact. Only with exactCommentCount
description	string or null	Full description
hashtags	array	Parsed from the description
description_links	array	Parsed from the description
keywords	array	The uploader's tags
is_short	boolean or null
is_live	boolean or null
available_languages	array	Every caption track the video has: code, name, is_generated
transcript_status	string	See the table above
transcript	string or null	Full text
transcript_chars	integer	Length, handy for filtering
transcript_segments	array	start, duration and text per line
transcript_language	string or null	The language actually returned
transcript_is_generated	boolean	Auto-generated, or human-written
_source	string	url, channel or keyword
_query	string	What produced this row

Sample row

{
  "video_id": "OrElyY7MFVs",
  "url": "https://www.youtube.com/watch?v=OrElyY7MFVs",
  "title": "My Evidence-Based Skincare Routine",
  "channel": "Ali Abdaal",
  "transcript_status": "ok",
  "transcript": "All right, so having the perfect skincare routine is actually pretty simple...",
  "transcript_chars": 18705,
  "transcript_language": "en",
  "transcript_is_generated": true,
  "transcript_segments": [
    { "start": 0.0, "duration": 0.88, "text": "All right, so having the perfect" }
  ],
  "_source": "url",
  "_query": "OrElyY7MFVs"
}

---

What to expect

Not every video has a transcript, and that is not a bug. Uploaders can switch captions off, and most Shorts have them off. In a 131-video sample across three keywords, 111 came back with a transcript; every one of the other 20 was disabled_by_uploader. With skipShorts on, a 39-video long-form sample returned 39.

You are only charged for transcripts actually returned. Rows with a non-ok status cost nothing.

Residential proxy is the default and worth keeping. YouTube rate-limits repeated requests from one IP, and a blocked request is retried on a fresh one.

Public content only. Private, unlisted and members-only videos are not accessible.

---

FAQ

Do I need a YouTube Data API key? No. No Google Cloud project, no quota, no login.

What if a video has no English transcript? You get whatever transcript it has, and transcript_language tells you which. A human-written track is preferred over an auto-generated one when both exist.

Can I get subtitles in SRT or VTT? The timed segments contain everything an SRT or VTT needs — start, duration and text — so converting is a few lines in your own code.

Does it translate? No. You get the transcript in the language it exists in.

Why is a transcript empty? Read transcript_status. disabled_by_uploader means the uploader turned captions off and no tool can retrieve it.

---

Our other Actors

• Threads Scraper — search Threads by keyword or pull a profile's posts, replies, reposts and media. No login.
• YouTube Monitor — watch keywords and channels for new videos only, with Shorts, exact publish times, like counts and built-in dedupe across runs.

---

Changelog

• 0.1.0 — First release. Videos, channels and keyword search; language fallback; transcript_status.

Disclaimer

This Actor collects publicly available captions only. You are responsible for how you use what you collect, including compliance with YouTube's terms of service and applicable law.

---

Keywords: youtube transcript, youtube transcript scraper, youtube captions, youtube subtitles, transcript api, video transcript, youtube to text, extract youtube transcript, bulk transcripts, transcript extractor, youtube captions api, srt, vtt, whisper alternative, no api key.