YouTube Transcript Scraper - Channels & Playlists
Pricing
Pay per event
Bulk YouTube transcripts from videos, whole channels and playlists: publish-date window, language priority with fallback, uploaded vs auto captions, video metadata and only-new-since-last-run.
Any mix of YouTube links and IDs, one per line: video links (watch, youtu.be, Shorts, live, embed) or 11-character video IDs; channel links (https://www.youtube.com/@TED, /channel/UC..., /c/..., /user/...) or bare @handles; playlist links (https://www.youtube.com/playlist?list=PL...) or playlist IDs. A channel link ending in /shorts or /streams reads that tab. A watch link that also carries &list= is read as that single video; use the /playlist?list= link for the whole playlist. If the list is empty, the Actor runs a small free example (3 videos from @TED) and says so in the status message.
[]Stop after this many transcripts in total across all sources. Only delivered transcripts count and are charged; rows for videos without captions are free and do not count.
How many videos to take from each channel or playlist (channels newest first, playlists in playlist order). Videos without captions count toward this limit; videos outside the publish-date window, without the title keywords or already delivered by an earlier run do not. 0 = no per-source limit (the run still stops at Maximum transcripts and reads at most 50 listing pages, about 5,000 videos, per source).
Which uploads to read from a channel: long-form videos (the channel's Videos tab), Shorts, live stream replays, or all uploads mixed. A channel link ending in /shorts or /streams overrides this for that channel. YouTube lists only the newest Shorts of a channel (measured: 100 of 520 for @TED); the run status says when a list is shorter than the channel's total.
Only videos published on or after this date (UTC). Use a date like 2026-09-01 or a period back from the run start like 7 days, 2 weeks, 3 months or 1 year. Checked against each video's exact publish date. On channels the listing stops once it reaches older videos. Leave empty for no lower limit.
Only videos published before this date (UTC, the day itself excluded), same format as above. Leave empty for no upper limit. On a channel the newer videos still have to be listed first, so a far-back window reads more listing pages.
Only videos whose title contains at least one of these words or phrases (case-insensitive substring match). Leave empty to take every video.
[]Caption language codes in order of preference, for example en, es, pt-BR. The first language the video has wins; en also matches en-US and en-GB, and pt-BR also matches pt. Leave empty to take the video's default caption track (usually its spoken language).
[]Uploaded (manual) captions are written by the channel; auto-generated captions come from YouTube speech recognition. Choose which to prefer within each language, or accept only one kind.
When a video has none of the caption languages above, return its default caption track instead (the row says isLanguageFallback: true). Turn off to get a free row with reason languageNotAvailable instead.
Add the caption lines with start time and duration in seconds (segments). The full plain text (transcriptText) is always included.
For scheduled runs. On a channel the Actor remembers every video an earlier run with the same settings read (with or without a transcript) and stops the listing (newest first) at the first of them, so each run returns only the uploads published since the last run; the first run returns the newest videos up to the per-channel limit. Videos in a temporary state are not remembered and are read again next time: upcoming premieres and scheduled live streams, empty caption files, YouTube bot checks, failed requests and videos without a publish date while a date window is set. On a channel they are read again only while they are newer than the stop point. On playlists and single videos, videos whose transcript was already delivered are skipped and videos without a transcript are tried again. The memory is kept per source and per combination of channel content, date window, title keywords, languages and caption settings (plus the optional memory name).
Optional name that keeps a separate only-new memory, for example one per schedule or per client. Runs with the same sources, settings and name share the memory.
YouTube is read through Apify Proxy (datacenter) by default. The Actor switches to a new proxy session when YouTube asks for a sign-in bot check, rate-limits, or returns an empty caption file. If a run fails with a block message, switch to the RESIDENTIAL group.
{ "useApifyProxy": true}