YouTube Transcript Scraper: Channels, Playlists & Shorts

YouTube Transcript Scraper: Channels, Playlists & Shorts

One rental for timestamped YouTube transcripts: single videos, Shorts, whole channels, playlists and keyword searches in one run, with the video's title, channel, views, duration and exact publish date in the same row.

No login, no API key, no browser. 256 MB. About 100 transcripts a minute.

Why this actor

	This actor	Typical transcript scrapers
Usage per 1,000 transcripts	~$0.13 (measured on Bytewells)	$2.00 – $5.00 per 1,000 elsewhere
Videos without captions	metadata row, or skipped	dropped
Channels, playlists, Shorts and search in one run	yes	often video URLs only
Exact publish date and time	yes	often relative ("3 weeks ago") or missing
SRT and WebVTT subtitles	yes	sometimes
Monitor channels for new uploads	built in (onlyNew)	do it yourself
Memory	256 MB	1 – 4 GB
Login or API key	no	sometimes

A video found by several inputs (a channel and a search, say) is delivered once.

Built for AI datasets and research

• LLM training and fine-tuning: spoken-language text with timestamps, language and whether captions are the creator's own or auto-generated.
• RAG and agents: index a channel's back catalogue, then keep it fresh with a daily only-new run.
• Summaries and chaptering: segments with start/end times map straight to clips and chapters.
• Content and market research: what a niche's top videos for a keyword actually say, with views and dates.
• Subtitles: SRT or WebVTT ready to download.

Export as JSON, CSV, Excel or XML, schedule runs, or deliver through webhooks and the Apify API.

What you get

One row per video:

Field
videoId, url, isShort
title, description, keywords, thumbnail
channelName, channelId, channelUrl
durationSeconds, viewCount
publishDate	ISO 8601 with time and zone, e.g. 2026-10-29T13:00:20-07:00
publishedText	as the channel or search page shows it ("2 weeks ago")
isLiveContent	a live stream or its replay
captionLanguages	every caption track: language, name, auto-generated or not
transcriptLanguage, transcriptAutoGenerated	the track that was read
hasTranscript	false when the video has no captions
transcript	[{start, end, text}], seconds
transcriptText	the whole transcript as one string
srt, vtt	when asked for
source, sourceUrl, scrapedAt	which input found the video

Example segment: {"start": 1.2, "end": 3.36, "text": "All right, so here we are, in front of the elephants"}

Quick start

Transcripts of a channel's 50 newest videos:

{
  "channels": ["@GoogleDevelopers"],
  "maxVideosPerSource": 50
}

Search results with SRT subtitles:

{
  "searches": ["sourdough starter"],
  "maxVideosPerSource": 30,
  "outputFormats": ["text", "srt"]
}

New uploads of several channels, every day (schedule it):

{
  "channels": ["@GoogleDevelopers", "@Google"],
  "channelTab": "all",
  "onlyNew": true,
  "monitorName": "google-channels"
}

Input

Input	Default	Notes
startUrls	—	videos, Shorts, youtu.be links, channels (/@handle, /videos, /shorts, /streams), playlists, search pages
channels	—	@handles, channel URLs or UC… ids
searches	—	keywords (videos, in YouTube's order)
channelTab	videos	videos, shorts, streams or all (videos + Shorts)
maxVideosPerSource	50	per channel tab, playlist or search
maxItems	1000	videos in total
transcripts	true	off: metadata only
language	en	e.g. de, es, pt-BR; en-GB also matches en; empty = the spoken language
fallbackToOriginalLanguage	true	no captions in your language: take the spoken-language transcript
preferManualCaptions	true	the creator's captions before auto-generated ones
includeSegments	true	the timestamped list
outputFormats	text	add srt and/or vtt
includePublishDate	true	exact date and time
skipVideosWithoutTranscript	false	leave caption-less videos out
onlyNew, monitorName	off	skip videos earlier runs delivered (remembered a year)
concurrency	4	videos read at once
proxyConfiguration	Apify Proxy	datacenter by default; switches to residential by itself if needed

Pricing

One monthly rental, no per-row fees: as many transcripts as you need. Runs use Bytewells compute from your credit, and this actor needs very little of it.

Measured on Bytewells (256 MB, datacenter proxy):

Run	Rows	Time	Usage
A channel (40) + a keyword search (40), with SRT	79 transcripts + 1 video without captions	59 s	$0.010 (~$0.13 / 1,000)

Videos YouTube doesn't show without signing in (age-restricted, private, removed) get a row with error.

Notes and limits

• About 1 in 8 videos has no captions at all (music, some live streams, short clips). They come with metadata only (hasTranscript: false), or not at all with skipVideosWithoutTranscript.
• Transcripts come from YouTube's captions: the creator's own, or YouTube's automatic speech recognition. Machine translation into other languages is not offered.
• With a language the video doesn't have, you get its spoken-language transcript and transcriptLanguage says which (turn off fallbackToOriginalLanguage to get none instead).
• Age-restricted, private and members-only videos can't be read without an account.
• Channel tabs are read newest first. With onlyNew, a channel stops at the first run of videos it already delivered, so daily runs take seconds.
• Searches follow YouTube's relevance order and end after about 300–350 videos per keyword (measured); use several related keywords for more.