Apify YouTube Scraper Alternatives: Free and Local (2026)

The Apify YouTube Scraper is one of the busiest actors on the platform: 167,601 installs, a 4.80/5 rating, and a meter that starts at $4 per 1,000 videos on the free plan. Its own output example ships with a comments: null field, because comment text is not what this actor sells. Comments live in a second paid actor and transcripts in a per-minute add-on, which is awkward, since those two are most of what people want when they go looking for an Apify YouTube scraper alternative.

The verdict first. Metadata on a list of video links: every route below is free, local, and I ran all of them this morning (October 16, 2026, home connection) against live videos. Comments: the official Data API is the only free route that worked in my test; yt-dlp’s comment collector returned zero on the current release. Transcripts: free for captioned videos, paid AI transcription only when a video has no captions at all. (This page is part of the web scraping tool alternatives series; each one runs what it recommends.)

Tool	The job it takes over	Runs where	The meter	Comments	Transcripts
Apify YouTube Scraper (baseline)	Search, channels, playlists, bulk video data	Apify cloud	$4 down to $2.40 per 1,000 videos; add-ons per video-minute	No, separate paid actor	Paid add-on, AI-generated
YouTube Data API v3	Comments, caption lists, search, official standing	Google’s cloud	Free under 10,000 quota units/day	Yes, 1 unit per call	Lists tracks; downloads only your own
yt-dlp	Everything about one video, locally	Your machine	None	Returned 0 in my test on the current release	Yes, captions to .vtt
youtube-transcript-api	Transcript text, nothing else	Your machine	None	No	Yes, about 1 second per video
Magpie	Six metadata fields per video link	Your machine	Free CLI	No	No

Ordering follows the jobs people bring to this actor, baseline first, then the tools by how much of the actor’s breadth each one absorbs.

What the Apify YouTube scraper does well

Search terms are first-class input. Paste a video, channel, playlist, or search-results link, or type keywords and cap the videos, Shorts, and streams separately. The README claims up to 20,000 videos per URL. Per video you get title, view count, likes, duration, upload date, hashtags, thumbnails, a monetization flag, and the description with its links parsed out. Per channel: name, subscriber count, location, joined date, total videos, total views. Exports run to JSON, CSV, Excel, and HTML, with Python and JavaScript SDKs, webhooks, and Make, Zapier, and Google Drive integrations. The store stats read 99.7% of runs succeeded.

Two holes sit in plain sight. The output schema carries commentsCount and commentsTurnedOff, but the comments field is null in the actor’s own example; comment text requires the separate YouTube Comments Scraper actor, on its own meter. Dislikes are gone entirely, removed because YouTube stopped publishing them. Transcription is a dropdown add-on: download existing subtitles, or pay per video-minute for speech-to-text AI on videos that have none, with an AI description and an AI summary as further per-minute extras.

One operational note from their own FAQ: a proxy is “essential” for this actor, datacenter recommended. Fair, since YouTube rate-limits cloud IPs harder than residential ones. But it means the per-1,000 price is a floor, not a ceiling, and your video list rides through someone else’s infrastructure on every run. The Apify alternatives overview covers the platform’s prepaid-credit mechanics in depth.

The meter, per 1,000 videos

The actor is priced per event: $0.004 per video on the Free plan, which is $4 per 1,000, dropping toward $2.40 per 1,000 as your plan discount deepens. The $5 of monthly free credits covers about 1,250 videos. Credits expire at the end of each month, and paid plans are prepayments: Starter is $19 that buys $19 of credits. A 10,000-video channel harvest costs $40 at the free-plan rate, which the free plan cannot actually fund past 1,250, or $24 at the discounted rate, before any transcription minutes.

yt-dlp: the free twin with an update treadmill

yt-dlp is the community’s answer to every YouTube layout change, released near-weekly (last push to the repo: October 10, 2026), 198,363 stars, Unlicense license. Metadata first, timed this morning:

yt-dlp --skip-download --no-warnings \
  --print "%(title)s | %(view_count)s | %(duration_string)s | %(channel)s" \
  "https://www.youtube.com/watch?v=jNQXAC9IVRw"

Me at the zoo | 444435061 | 19 | jawed

2.0 seconds, and the view count matches what Magpie’s extractor returned for the same video, to the digit.

Subtitles work the same way. One command wrote a 4.3 KB English .vtt for the Rick Astley video in 3.4 seconds:

yt-dlp --skip-download --no-warnings --write-subs --write-auto-subs \
  --sub-langs en --sub-format vtt -o "captions" \
  "https://www.youtube.com/watch?v=dQw4w9WgXcQ"

Comments are where it fell over. On yt-dlp 2026.08.19, the current release as I write, bounded --write-comments runs collected 0 comments on two videos whose comment_count fields read 2.4 million and 10 million (the latter is YouTube’s display cap). An unbounded run passed five minutes and I killed it. That is the treadmill in one incident: YouTube changes the comments plumbing, the fixed release ships within days, and your pipeline works again after pip install -U yt-dlp. When comment text is the core deliverable, I would not bet on this route without a monitored updater. The Data API is the stable free one.

YouTube Data API: the official route for comments

Setup is a Google Cloud project, the API enabled, and a key. The quota, verified against Google’s own calculator page this week: 10,000 units per day across endpoints, with search.list in a separate bucket of 100 calls per day, resetting at midnight Pacific. The units go far for this actor’s core job. videos.list costs 1 unit and takes up to 50 video IDs per call, so the free tier covers half a million metadata reads a day. commentThreads.list is also 1 unit. captions.list is 50.

curl "https://www.googleapis.com/youtube/v3/videos?part=snippet,statistics&id=jNQXAC9IVRw&key=YOUR_KEY"

(That shape is documented API, not a live run; I hold no key on this machine and won’t invent output.) One real limit: captions.download works only for videos you own, which is exactly why the next entry exists.

youtube-transcript-api: transcripts for free

A short Python call returns the transcript of any captioned video. MIT license, 8,466 stars, version 1.2.4 installed here today:

from youtube_transcript_api import YouTubeTranscriptApi

t = YouTubeTranscriptApi().fetch("jNQXAC9IVRw")
print(len(t.snippets), t.language_code)
print(t.snippets[0].text)

6 en
All right, so here we are, in front of the
elephants

That is “Me at the zoo”: six segments in 1.2 seconds. The Rick Astley video gave 61 segments in 1.0 second. What it does not do: metadata beyond the transcript, comments, search, or videos whose owners disabled captions. Caption-less audio is the actor’s AI add-on territory, or a local speech-to-text model such as Whisper, which needs the audio file first.

Where Magpie fits, stated plainly

Magpie’s youtube vertical returns six fields per video link: title, author, full description, view count, duration in seconds, canonical URL. No API key, no per-record cost, one fetch per video. This morning’s run took 2.4 seconds:

magpie vertical "https://www.youtube.com/watch?v=jNQXAC9IVRw" --name youtube

{
  "record": {
    "author": "jawed",
    "duration_s": 19,
    "title": "Me at the zoo",
    "url": "https://www.youtube.com/watch?v=jNQXAC9IVRw",
    "views": 444435061
  }
}

(Abridged; the full description field runs ten lines.) The limit, stated now so you don’t find it after building on this: no comments, no subtitles or transcripts, no channels, playlists, or search. If your job is comment text or caption files, this is not your tool, and no flag changes that. Where it fits is bulk metadata: a list of links through a loop lands in CSV at about two seconds per video, and the YouTube scraper guide carries the tested pipeline with its failure modes. For the spreadsheet-shaped general case, Scrape a Website to CSV or JSON covers headers, dedupe, and Excel. The CLI is free, licensed AGPL-3.0-or-later or commercial; the desktop app allows 5 free runs per install, then $49 a month.

Pick by the job

• Metadata for a pile of links, zero budget: Magpie’s vertical or yt-dlp. Both free, both about 2 seconds per video, and they agree on the numbers.
• Comment text: the Data API at 1 unit per call, or the actor’s separate comments actor if you want cloud scale and accept the second meter.
• Transcripts: youtube-transcript-api for captioned videos, yt-dlp for .vtt files, and the actor’s AI transcription only when captions don’t exist.
• Search results or channel walks at the actor’s 20,000-per-URL scale: the one job where its breadth has no free twin. The API’s search bucket is 100 calls a day and the local tools refuse search entirely.

FAQ

Is there a free alternative to the Apify YouTube Scraper?

For per-video metadata and transcripts of captioned videos, yes: yt-dlp, youtube-transcript-api, and Magpie’s vertical all ran free today. Nothing free replaces channel walks and search harvests at scale; the API’s search bucket is 100 calls a day.

Does the Apify YouTube Scraper include comments?

No. The actor’s own output example shows a commentsCount field and a null comments field. Comment text comes from the separate YouTube Comments Scraper actor, billed on its own meter.

How do I get YouTube transcripts without paying?

youtube-transcript-api returns caption text free, about a second per video. yt-dlp’s --write-subs writes .vtt files. The actor’s per-minute AI transcription matters only when a video has no captions at all.

Is scraping YouTube legal?

Two layers. robots.txt permits watch-page crawls, but YouTube’s Terms of Service restrict automated access as a contract matter. Public metadata at low volume is common practice; republishing personal data is where people get hurt. Not legal advice; Is Web Scraping Legal? has the plain version.

My own setup: the vertical loop for metadata, youtube-transcript-api when I need the words, the Data API if comments ever matter, and the actor only when a job arrives with 20,000 links or a search term in it. The social media scraper roundup ranks what’s collectible across platforms if YouTube is one stop of several.