Do I need a YouTube API key to scrape videos?

No. Videos, channels, playlists, transcripts and comments are public pages. A scraper reads them the way a visitor without an account does, so there is no key, no quota and no Google Cloud project.

How do I get YouTube transcripts in bulk?

Run the scraper with transcripts on: it returns each video's subtitles as timed segments and as one plain text, manual captions preferred and auto-generated ones as fallback, in the language you choose when it exists. Works on a search, a channel, a playlist or a list of URLs.

What does the YouTube Data API quota allow?

10,000 units a day per project. A search costs 100 units, so about 100 searches a day; video details cost 1 unit each. Transcripts are not available through the Data API at all, and comments need extra permissions.

What does the scraper cost?

$0.50 per 1,000 rows, a row being a video, a comment or a reply; $1 per 1,000 videos with a transcript; $2 per 1,000 channel records. No start fee, platform usage and proxies included.

YouTube scraper and API: videos, transcripts and comments without an API key

The YouTube Data API is free, official and small: 10,000 units a day, a search costs 100 of them, transcripts are not offered at all. Everything on a public video page is visible without an account, and that is what a scraper reads. This page shows what you get per video, how transcripts and comments work, where the limits are, and what it costs.

Updated September 2026. Fields and limits measured on the live site.

Data API versus scraper

	YouTube Data API v3	Scraper
Setup	Google Cloud project, API key, OAuth for some calls	a keyword or a URL
Quota	10,000 units a day; search = 100 units, video = 1, comment thread page = 1	none, pay per row
Transcripts	not available (captions download needs the owner's OAuth)	yes, any public video with captions
Comments	yes, 100 per page	yes, top or newest
Likes, tags, category	yes	yes
Subscriber and comment counts	yes	yes, as YouTube shows them (rounded)
Channel record (country, joined date, links)	partly (no links)	yes
Best for	an app that acts on behalf of users, small volumes	research, monitoring, datasets, AI pipelines

The scraper

The YouTube Scraper on Apify takes search queries, channel URLs, playlist URLs or video URLs and returns one row per video. Transcripts and comments are options on the same run. A row carries:

Video: id, url, title, durationSeconds, viewCount, likeCount, publishedAt (date and time), category, description, keywords, thumbnailUrl, isShort, isLive.
Channel: channelName, channelId, channelUrl, subscriberCount; commentCount with the counts option.
Transcript: transcript as timed segments {start, duration, text}, transcriptText as one string, transcriptLanguage.
Comments: separate rows with text, author, authorChannelId, likeCount, replyCount, publishedText and the videoId they belong to; replies as rows under their comment with parentId.
Channel record (option): one row per channel with description, country, joinedAt, all-time viewCount, videoCount, subscriberCount, handle and external links.

Speed: about 100 videos with details in 10-60 seconds depending on the source; listing only, several hundred a minute.

Calling it from code

curl -X POST "https://api.apify.com/v2/acts/alfalfa~youtube-scraper/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H 'Content-Type: application/json' \
  -d '{"searchQueries":["python tutorial"],"sortBy":"views","uploadDate":"month","maxVideos":50,"transcript":true}'

Longer jobs: start the run, poll it, read the dataset when it finishes. Schedules, webhooks and the n8n, Make and Zapier nodes work as for any Apify Actor; the API overview shows the pattern.

Things to know before you trust the numbers

Search is what a visitor sees. About 20 videos per page, rarely more than 500-600 results for one query, US English settings. Split by upload date for more.
Channels and playlists go deep. A channel's videos tab or its uploads playlist pages through the whole history, thousands of videos.
Rounded counts. Views and likes are exact; subscriber and comment counts are what YouTube displays ("21.2M", "2.4M").
Transcripts need captions. Manual or auto-generated; live streams have none while live. Multi-hour videos produce tens of thousands of segments.
Comment replies come on request, up to a number per comment, and count towards the per-video comment limit.

Uses that pay for themselves

AI and research. Transcripts of a whole channel or a topic as plain text for summaries, RAG or translation.
Monitoring. Every new video on a keyword this week, with views and likes, into a sheet; a competitor's channel every Monday.
Comment analysis. Top or newest comments of a launch video or a podcast with likes and authors, for sentiment and feedback mining.
Content planning. The most viewed videos of a niche this month, their length, tags and titles.

Questions people ask

Is this allowed?

The scraper reads public pages without logging in and stores nothing but the results in your dataset. Respect YouTube's terms and the uploader's copyright when you reuse content, and personal data laws for comment authors.

Why is it this cheap?

A video's details are a few kilobytes and the scraper runs in a few megabytes of memory, so platform costs per thousand videos are cents. Transcripts are bigger, hence their own small price.

Does it download videos?

No. Metadata, transcripts and comments only.

Open the YouTube Scraper Read: the API overview