🎬 YouTube Transcript Scraper — Bulk YouTube Transcripts, Subtitles & Captions API

YouTube Transcript Scraper is a fast, reliable Apify Actor for YouTube transcript extraction, YouTube subtitle scraping, and caption download at scale. Paste one video link or ten thousand — plus playlists, channels, Shorts, or just a search keyword — and get back clean, structured transcripts together with the video details you care about: channel name, channel ID, publish date, view count, likes, comments, keywords, thumbnail and description.

Every video is written to your Apify Dataset the moment its transcript is ready, so you can export to JSON, CSV, Excel, or stream rows straight into RAG pipelines, LLM knowledge bases, vector databases, content research, and SEO analysis while the run is still going.

💡 Need a custom version, a private integration, or a tailored scraper? Email hello.dataminds@gmail.com.

---

📑 Table of contents

• What is YouTube Transcript Scraper?
• Main features
• Who is this Actor for?
• Built for generative AI, LLMs, and RAG
• What the scraper does
• Supported YouTube link types
• Transcript formats
• Input
• Output format (Dataset)
• Example output (JSON)
• Quick start
• Input parameters reference
• Integrations: LangChain, LlamaIndex, OpenAI, Pinecone, n8n
• Pricing & how to control cost
• Frequently asked questions (FAQ)
• Troubleshooting
• Help, support & custom builds
• Is scraping YouTube transcripts legal?
• SEO keywords targeted

---

🎬 What is YouTube Transcript Scraper?

YouTube Transcript Scraper is a hosted YouTube transcript API that turns spoken video content into clean, machine-readable text. Unlike browser extensions or one-video-at-a-time tools, this Actor:

• Handles bulk input — a list of URLs, an uploaded file, a Google Sheet, or a whole channel.
• Reads both caption types — human-written subtitles and auto-generated captions, preferring the better one.
• Speaks 76 languages with a smart fallback chain, so you get a transcript instead of an empty row.
• Returns four formats — a clean text array, one continuous string, ready-to-use WebVTT, or ready-to-use SRT.
• Adds video metadata — views, likes, comments, tags, thumbnail, description, channel and publish dates.
• Streams results live into an Apify Dataset, exportable to JSON / CSV / Excel / XML or read via the Apify API.

If you have ever needed to "turn these 500 YouTube videos into text my model can read", this is the Actor.

---

✨ Main features

• 🔗 True bulk input — paste a list, upload a file, or link a Google Sheet. One URL or ten thousand, same run.
• 📺 Channels & playlists auto-expanded — drop in youtube.com/@handle or a playlist link and the Actor finds the videos for you.
• 🔍 Search-keyword mode — no links at all? Type apify tutorial and get the top matching videos' transcripts.
• 📱 Shorts support — include or exclude YouTube Shorts with one toggle; a Shorts link you paste yourself always works.
• 🧾 Four transcript formats — text array (best for AI), single-line text, WebVTT, and SubRip / SRT with real timings.
• 🌍 76 caption languages with automatic fallback: your language → English → whatever the video offers.
• 🗣️ Manual and auto-generated captions — human-written subtitles are preferred, auto captions used when that's all there is.
• 🎁 Eleven optional data points — tick only what you need, keep rows lean and runs cheap.
• 🛡️ Built-in anti-block handling — the run adapts automatically when YouTube pushes back, so large batches keep flowing without you configuring anything.
• 🔁 Configurable retries — 3 to 12 attempts per video with exponential backoff and jitter.
• ⚡ Parallel processing — 1 to 16 videos at a time, with a politeness delay you control.
• 💾 Real-time Dataset writes — a row appears the instant its transcript is ready; an aborted run still keeps everything collected so far.
• 📊 Five prebuilt Output views — Quick View, Transcript, Channel & Publishing, Engagement Stats, Metadata & Media.
• 🧾 Skipped-video audit — every video that could not be collected is written to skipped-videos.json with the exact reason.
• ✂️ Oversized-row protection — a six-hour video's transcript is trimmed to fit instead of failing the row.
• 🧹 Automatic de-duplication — the same video reached from two different links is only processed once.

---

👥 Who is this Actor for?

• 🤖 AI engineers building RAG, retrieval-augmented generation, and LLM grounding pipelines from video.
• ✍️ Content marketers & SEO specialists repurposing videos into blog posts, newsletters and show notes.
• 📈 Competitive analysts mining competitor channels for topics, positioning, and messaging.
• 🎧 Podcasters & video creators generating searchable archives, chapters, and subtitles.
• 🔬 Researchers collecting spoken-language corpora for NLP, sentiment, and discourse analysis.
• ♿ Accessibility teams exporting SRT / VTT subtitle files for captioning workflows.
• 📚 Educators & students turning lectures and tutorials into study notes.
• 🛠️ Developers who want a YouTube transcript API they can call from any language.

---

🤖 Built for generative AI, LLMs, and RAG

Video is the largest untapped text corpus on the web. This Actor makes it usable:

• RAG pipelines — chunk captions and embed them alongside title + channelName metadata.
• LLM knowledge bases — ground chatbots in webinars, tutorials, conference talks and product demos.
• Semantic video search — index thousands of videos so users can search what was said, not just titles.
• Vector database ingestion — drop straight into Pinecone, Qdrant, Weaviate, Chroma, Milvus, or pgvector.
• Custom GPT / Claude Project knowledge files — export the Dataset JSON and upload.
• Summarization & repurposing — feed a transcript to an LLM and get a blog post, thread, or newsletter.
• Content moderation & compliance — scan spoken content at scale.

Practical AI use cases

1. Turn a whole YouTube channel into a chatbot knowledge base — paste the channel URL, set a limit, embed the output.
2. Auto-generate blog posts from webinars — single-line text format straight into your writing prompt.
3. Build a searchable conference-talk archive — transcript + speaker channel + publish date per row.
4. Competitive messaging analysis — pull the last 100 videos from three competitor channels and diff the language.
5. Subtitle production — export the SRT format and ship the files.
6. Lecture-to-notes pipelines — playlist in, structured study notes out.
7. Trend detection — search-keyword mode on a schedule, then topic-model the transcripts.

---

⚙️ What the scraper does

• Reads every link you give it, separates individual videos from playlists and channels, and expands the latter into video lists (up to Max videos per source).
• Runs your search keywords and adds the top matching videos to the same queue.
• De-duplicates the whole queue so nothing is collected twice.
• Pulls the best available caption track for each video — your preferred language first, then English, then whatever exists; human-written subtitles preferred over auto-generated ones.
• Renders the transcript in the format you chose and attaches the optional data points you ticked.
• Writes each row to the Apify Dataset immediately — downstream systems can consume rows mid-run.
• Keeps going when a video fails — private, deleted, region-locked and caption-less videos are recorded in a skip report, never abort the run.

---

🔗 Supported YouTube link types

Link type	Example	Behaviour
Video	https://www.youtube.com/watch?v=Qb9s3UiMSTA	Processed directly
Short link	https://youtu.be/Qb9s3UiMSTA	Processed directly
Shorts	https://www.youtube.com/shorts/VIDEO_ID	Processed directly
Live replay	https://www.youtube.com/live/VIDEO_ID	Processed directly
Playlist	https://www.youtube.com/playlist?list=PL...	Expanded, up to Max videos per source
Channel	https://www.youtube.com/@handle	Expanded, up to Max videos per source
Channel tab	https://www.youtube.com/@handle/videos	Expanded, up to Max videos per source
Search keyword	apify web scraping (in Search keywords)	Top results expanded

ℹ️ Non-YouTube links are skipped with a clear warning instead of failing the run.

---

🧾 Transcript formats

outputFormat	captions value	Best for
json (default)	Array of caption lines — ["line one", "line two", …]	AI, search, analytics, spreadsheets
text	One continuous, tag-free string	Summarization, LLM prompts, blog drafts
vtt	Full WebVTT document with 00:00:01.360 --> 00:00:03.040 timings	Web players, <track> elements
srt	Full SubRip document with numbered cues	Video editors, subtitle files

---

📥 Input

Only one thing is really required: either at least one YouTube URL or at least one search keyword.

{
  "urls": [
    "https://www.youtube.com/watch?v=Qb9s3UiMSTA",
    "https://www.youtube.com/@Apify"
  ],
  "searchKeywords": ["web scraping tutorial"],
  "scrapeShorts": false,
  "maxVideosPerSource": 10,
  "maxResults": 0,
  "outputFormat": "json",
  "language": "en",
  "channelNameBoolean": true,
  "channelIDBoolean": true,
  "datePublishedBoolean": true,
  "dateTextBoolean": true,
  "relativeDateTextBoolean": true,
  "viewCountBoolean": true,
  "likesBoolean": true,
  "commentsBoolean": true,
  "keywordsBoolean": true,
  "thumbnailBoolean": true,
  "descriptionBoolean": true,
  "maxConcurrency": 4,
  "requestDelaySecs": 1.0,
  "maxRetries": 8
}

---

📤 Output format (Dataset)

Results land in the default Apify Dataset. You can:

• Browse rows in the Apify Console with five prebuilt views — each section of the data on its own tab.
• Export to JSON, CSV, Excel, XML, HTML, or RSS.
• Pull data programmatically via the Apify API.
• Stream into AI, ETL, or analytics pipelines while the run is still active.

Output views (each section of the table, separately)

View	Columns
✨ Quick View	videoId, title, channelName, datePublished, viewCount, captions
📝 Transcript	videoId, title, captions
📺 Channel & Publishing	videoId, title, channelName, channelID, datePublished, dateText, relativeDateText
📈 Engagement Stats	videoId, title, viewCount, likes, comments
🏷️ Metadata & Media	videoId, title, keywords, thumbnailUrl, description

Row fields

Field	Always present?	Description
videoId	✅ Always	The YouTube video ID, e.g. Qb9s3UiMSTA.
title	✅ Always	The video title.
channelName	Optional	Channel the video was published on. Enable 📺 Channel Name.
channelID	Optional	The channel's unique ID. Enable 🆔 Channel ID.
datePublished	Optional	Full ISO publish timestamp, e.g. 2024-02-22T23:00:12.000Z.
dateText	Optional	Friendly date, e.g. Feb 22, 2024.
relativeDateText	Optional	Human age, e.g. 2 years ago.
viewCount	Optional	Formatted views, e.g. 27,406,159 views.
likes	Optional	Formatted like count, e.g. 564,840.
comments	Optional	Formatted comment count, e.g. 30,915.
keywords	Optional	Video tags as a comma-separated string.
thumbnailUrl	Optional	Direct link to the video thumbnail.
description	Optional	Full video description text.
captions	✅ Always	The transcript, in the format you selected.

🧾 Videos that could not be collected are written to skipped-videos.json in the run's key-value store, each with a reason (No transcript available for this video, Private video, Transcript download refused (HTTP 429), …).

---

🧪 Example output (JSON)

Real output for https://www.youtube.com/watch?v=Qb9s3UiMSTA with every optional data point enabled (transcript abbreviated — the real row carries all 1,093 caption lines):

{
  "videoId": "Qb9s3UiMSTA",
  "title": "Asyncio in Python - Full Tutorial",
  "channelName": "Tech With Tim",
  "channelID": "UC4JX40jDee_tINbkjycV4Sg",
  "datePublished": "2024-04-05T14:30:26.000Z",
  "dateText": "Apr 5, 2024",
  "relativeDateText": "2 years ago",
  "viewCount": "354,405 views",
  "likes": "9,871",
  "comments": "190",
  "keywords": "tech with tim",
  "thumbnailUrl": "https://i.ytimg.com/vi/Qb9s3UiMSTA/maxresdefault.jpg",
  "description": "Asynchronous programming allows our code to be more efficient by doing multiple things at once without any unnecessary waiting …",
  "captions": [
    "Imagine programing",
    "is a journey from point A to D.",
    "In traditional synchronous programing",
    "we travel in a straight line",
    "stopping at each point"
  ]
}

The same video with "outputFormat": "srt":

{
  "videoId": "Qb9s3UiMSTA",
  "title": "Asyncio in Python - Full Tutorial",
  "captions": "1\n00:00:00,033 --> 00:00:01,133\nImagine programing\n\n2\n00:00:01,133 --> 00:00:03,466\nis a journey from point A to D."
}

---

🚀 Quick start

Run in Apify Console

1. Log in at console.apify.com → Actors.
2. Open YouTube Transcript Scraper.
3. Paste your 🔗 YouTube URLs (or type a 🔍 search keyword).
4. Pick a 🧾 Transcript format and a 🌍 caption language.
5. Tick the 🎁 extra data points you want as columns.
6. Click Start and watch rows stream into the Output tab in real time.
7. Export as JSON / CSV / Excel, or pull via the Apify API.

Run via API

curl -X POST "https://api.apify.com/v2/acts/<your-actor-id>/runs?token=<APIFY_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{
        "urls": ["https://www.youtube.com/watch?v=Qb9s3UiMSTA"],
        "outputFormat": "json",
        "language": "en",
        "channelNameBoolean": true,
        "viewCountBoolean": true
      }'

Run it synchronously and get the rows back in one call:

curl -X POST "https://api.apify.com/v2/acts/<your-actor-id>/run-sync-get-dataset-items?token=<APIFY_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{ "urls": ["https://youtu.be/Qb9s3UiMSTA"] }'

Read the rows from an existing run:

$curl "https://api.apify.com/v2/actor-runs/<runId>/dataset/items?format=json&token=<APIFY_TOKEN>"

Run from Python

from apify_client import ApifyClient

client = ApifyClient("<APIFY_TOKEN>")
run = client.actor("<your-actor-id>").call(run_input={
    "urls": ["https://www.youtube.com/@Apify"],
    "maxVideosPerSource": 25,
    "outputFormat": "text",
    "channelNameBoolean": True,
})

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["title"], "→", item["captions"][:120])

---

🛠️ Input parameters reference

Parameter	Type	Default	Description
urls	array	[]	YouTube links. Videos, Shorts, youtu.be links, playlists, channels. Bulk-friendly: list, file upload, or Google Sheet.
searchKeywords	array	[]	Search terms; the top matching videos are added to the queue. Use instead of, or on top of, urls.
scrapeShorts	bool	false	Include Shorts (≤ 60 s) found inside channels, playlists and searches. Shorts you paste directly are always processed.
maxVideosPerSource	int	10	Videos taken from each playlist, channel, or keyword. Does not limit individual video links.
maxResults	int	0	Hard ceiling on videos processed in the whole run, after expansion and de-duplication. 0 = no limit.
outputFormat	enum	json	json (text array) · text (single string) · vtt (WebVTT) · srt (SubRip).
language	enum	en	Preferred caption language (76 options). Falls back to English, then any available track.
channelNameBoolean	bool	false	Add channelName.
channelIDBoolean	bool	false	Add channelID.
datePublishedBoolean	bool	false	Add datePublished (ISO timestamp).
dateTextBoolean	bool	false	Add dateText (Feb 22, 2024).
relativeDateTextBoolean	bool	false	Add relativeDateText (2 years ago).
viewCountBoolean	bool	false	Add viewCount.
likesBoolean	bool	false	Add likes.
commentsBoolean	bool	false	Add comments.
keywordsBoolean	bool	false	Add keywords (tags).
thumbnailBoolean	bool	false	Add thumbnailUrl.
descriptionBoolean	bool	false	Add description.
maxConcurrency	int	4	Videos processed in parallel (1–16).
requestDelaySecs	number	1.0	Polite pause before each request starts. Decimals allowed; 0 disables it.
maxRetries	int	8	Attempts per video before it is skipped (3–12), with exponential backoff.

---

🔌 Integrations: LangChain, LlamaIndex, OpenAI, Pinecone, n8n

• 🦜 LangChain — load the Dataset with ApifyDatasetLoader, mapping captions → page_content.
• 🦙 LlamaIndex — feed transcripts into a VectorStoreIndex with title/channelName metadata.
• 🧠 OpenAI — chunk and embed with text-embedding-3, or upload as a Custom GPT knowledge file.
• 🤖 Anthropic Claude — attach the Dataset JSON to Claude Projects or the Files API.
• 📍 Pinecone, Qdrant, Weaviate, Chroma, Milvus, pgvector — embed and upsert with metadata filters.
• 📊 n8n, Make.com, Zapier — trigger runs and route transcripts into your automations.
• 📦 Snowflake, BigQuery, PostgreSQL, MongoDB, S3 — load Dataset JSON/CSV into your warehouse.
• 🔔 Webhooks — get notified the moment a run finishes; see Apify integrations.

---

💸 Pricing & how to control cost

Billable event

Event	When it fires	What you get
transcript_result	Once per transcript row written to the Dataset	A complete video record — transcript plus every data point you enabled

Videos that fail, have no captions, are private/deleted, or are dropped by maxResults are not charged. If a run hits its charging limit (ACTOR_MAX_TOTAL_CHARGE_USD), it stops cleanly and everything already collected stays in your Dataset.

⚙️ Publishing note: the transcript_result event must exist in Apify Console → your Actor → Publication → Monetization with the same name. Until pay-per-event is configured the Actor still runs and saves data normally.

Keeping runs cheap

• 🎯 Set maxResults and maxVideosPerSource — the single biggest cost lever.
• 🎁 Only tick the extra data points you actually need.
• 🧾 Use json or text output rather than vtt/srt when you don't need timings — smaller rows.
• ⚡ Leave maxConcurrency at 4 unless you have a reason to change it.
• 🐢 Raise requestDelaySecs and lower maxConcurrency on very large batches — a gentler run finishes more videos on the first try.

---

❓ Frequently asked questions (FAQ)

Do I need a YouTube API key? No. There is nothing to configure beyond the Actor input.

Does it work on videos without subtitles? It works on any video that has some caption track — human-written or auto-generated. Videos with captions fully disabled by the uploader cannot be transcribed and are listed in skipped-videos.json.

Can I scrape a whole channel? Yes. Paste https://www.youtube.com/@handle and set Max videos per source.

Can I get transcripts without giving any URLs? Yes — use 🔍 Search keywords.

Which languages are supported? 76 caption languages are selectable. If your language isn't available on a given video, the Actor falls back to English and then to any track that exists, so you still get a transcript.

Can I download SRT or VTT subtitle files? Yes. Choose srt or vtt as the transcript format; the captions field then contains a complete, ready-to-save subtitle document.

Are YouTube Shorts supported? Yes. A Shorts link you paste is always processed. To also pull Shorts found inside channels/playlists/searches, turn on 📱 Include YouTube Shorts.

How fast is it? Typically 1–3 seconds per video, processed 4 at a time by default. A 100-video channel usually finishes in a couple of minutes.

What happens if a video fails? The run continues. The failure and its reason are recorded in skipped-videos.json, and the summary at the end of the log shows how many videos were skipped.

Can I run it on a schedule? Yes — use Apify Schedules, or trigger it from n8n / Make / your own backend.

Does an aborted run lose my data? No. Rows are written to the Dataset as they are produced, so everything collected before the abort is already saved.

---

🛠️ Troubleshooting

Symptom	Likely cause	Fix
No transcript available for this video	The uploader disabled captions entirely.	Nothing to fix — the video genuinely has no caption track.
Transcript is in the wrong language	The video has no track in your chosen language.	Set 🌍 Preferred caption language to a language the video actually offers, or leave it on en.
Private video / Video unavailable	The video is private, deleted, or region-locked.	Nothing to fix — use a different link.
Transcript download refused (HTTP 429)	Rate limiting.	The Actor recovers from this on its own. To reduce it further, lower maxConcurrency and raise requestDelaySecs.
Channel returned fewer videos than expected	maxVideosPerSource or maxResults capped it.	Raise those limits.
Shorts missing from a channel run	scrapeShorts is off.	Turn on 📱 Include YouTube Shorts.
Run saved 0 rows	No valid YouTube links, or none of the videos have captions.	Check the log — every skipped video is listed with its reason.

---

🤝 Help, support & custom builds

• 💬 Issues & feature requests — use the Issues tab on the Actor page in Apify Console.
• 📧 Direct support, custom scrapers & private integrations — hello.dataminds@gmail.com.
• 📚 Platform docs — Apify documentation.

We reply to every message. If you need a variant of this Actor tuned to your workflow, just ask.

---

⚖️ Is scraping YouTube transcripts legal?

This Actor collects only publicly available information — captions and metadata that YouTube already serves to any visitor. It does not access private videos, members-only content, or anything behind a login.

That said, you are responsible for how you use the data. Respect copyright, YouTube's Terms of Service, and applicable law (GDPR, CCPA, and similar). Transcripts of copyrighted videos remain the property of their creators — use them for research, analysis, accessibility, or internal purposes, and seek permission before republishing. If you're unsure, talk to a lawyer.

See Apify's take: Is web scraping legal?

---

🔍 SEO keywords targeted

youtube transcript scraper · youtube transcript api · youtube subtitle scraper · youtube caption downloader · bulk youtube transcript extraction · download youtube transcript · youtube transcript to text · extract youtube subtitles · youtube channel transcript scraper · youtube playlist transcript · youtube shorts transcript · youtube srt download · youtube vtt download · auto generated captions scraper · youtube video to text · transcribe youtube video · youtube transcript for AI · youtube data for RAG · youtube scraper apify · video transcript dataset · youtube metadata scraper · youtube views likes comments scraper · multilingual youtube transcripts · youtube transcript json · youtube subtitles api · youtube content research tool