YouTube Transcript Downloader & Extractor: Subtitles to JSON

Download a YouTube transcript as structured JSON from a video URL or video id. Paste one link or a long list, choose preferred languages, keep manual subtitles when available, and optionally include timestamps for every subtitle segment. You receive one dataset row for each nonblank input, in the same order, ready for AI pipelines, search indexes, research notes or a spreadsheet.

The actor reads the captions available for the video. It supports manual subtitles and automatically generated subtitles. It does not create a transcript from audio, translate a video, or invent missing subtitles. When a video cannot provide subtitles, its result explains the failure and carries no transcript charge.

Input

videoUrls is required and accepts a list of strings. Each string may be a standard YouTube watch URL, a shortened youtu.be URL, a mobile watch URL, a shorts URL, an embed URL, a /v/ URL, or an eleven-character video id. Surrounding spaces are removed. Blank entries are skipped; if every entry is blank, the run produces no dataset rows and records an input error.

This input was used in an actual run:

{
  "videoUrls": [
    "https://youtu.be/jNQXAC9IVRw",
    "8jPQjjsBbIc",
    "https://www.youtube.com/watch?v=UF8uR6Z6KLc&t=10s"
  ]
}

Field	Default	Meaning
videoUrls	Required	Video URLs or ids to process
preferred_languages	["en"]	Preferred subtitle language codes, in priority order
prefer_manual	true	Prefer requested manual tracks before automatic tracks
include_segments	false	Include every segment with its start and duration
concurrency	10	Requested concurrency from 1 to 50

You may submit the same video more than once. Each occurrence remains a separate result. The order of the returned rows follows the input order, even when videos finish at different times. A start-time parameter in a watch URL does not trim the transcript; the result covers the complete selected subtitle track.

Language selection

With prefer_manual enabled, the actor first checks all requested languages for manual subtitles, then checks those languages for automatic subtitles. With it disabled, each requested language is checked in order, with manual subtitles still preferred within that language. If none of the requested languages is available, the first available manual track is selected, followed by an automatic track when there are no manual tracks.

Language matching ignores letter case. For example, pt-BR selects an available Brazilian Portuguese track and the result keeps YouTube's original language code. A regional language code still differs from its shorter counterpart: en-US is not treated as en. The actual selected language always appears on the result, so check that field when fallback matters to your application.

YouTube transcript output

This is a real row from the run above, for the input https://youtu.be/jNQXAC9IVRw:

{
  "input": "https://youtu.be/jNQXAC9IVRw",
  "video_id": "jNQXAC9IVRw",
  "fetched_at": "2026-10-02T02:13:46.407Z",
  "video_url": "https://www.youtube.com/watch?v=jNQXAC9IVRw",
  "source": "youtube-transcript",
  "platform": "youtube",
  "language": "en",
  "is_generated": false,
  "segments": 6,
  "segment_count": 6,
  "word_count": 39,
  "duration_seconds": 18.88,
  "transcript": "All right, so here we are, in front of the elephants the cool thing about these guys is that they have really... really really long trunks and that's cool (baaaaaaaaaaahhh!!) and that's pretty much all there is to say",
  "text": "All right, so here we are, in front of the elephants the cool thing about these guys is that they have really... really really long trunks and that's cool (baaaaaaaaaaahhh!!) and that's pretty much all there is to say"
}

With include_segments set to true and German preferred, the same video returned these timed segments in an actual run:

"segments_data": [
  {"start": 0, "duration": 5.07, "text": "Also hier sind wir vor den Elefanten."},
  {"start": 5.07, "duration": 7.4, "text": "Das Coole an den Typen ist dass sie sehr, sehr, sehr, lange Rüssel haben."},
  {"start": 12.47, "duration": 1.8, "text": "Und das ist cool."},
  {"start": 17.07, "duration": 3.93, "text": "Ansonsten gibt es nicht wirklich viel zu sagen."}
]

Successful rows contain these established fields:

Field	Type	Meaning
input	string	Original input after trimming
video_id	string	Extracted video id
video_url	string	Canonical watch URL
language	string	Selected subtitle language
is_generated	boolean	Whether subtitles were generated automatically
segments	integer	Number of subtitle segments
word_count	integer	Word count in the joined text
duration_seconds	number	End time of the last subtitle segment
transcript	string	Complete joined subtitle text
segments_data	array	Segment objects, only when requested

Each segment object has start, duration, and text. Times are measured in seconds and rounded to two decimal places. Segment text retains line breaks, while the joined transcript normalizes whitespace. HTML entities are decoded and formatting tags are removed.

Additional fields make the rows easier to combine with other data: source, platform, fetched_at, text, and segment_count. text contains the same content as transcript; segment_count contains the same count as segments. Existing fields retain their names and types.

The run summary counts all dataset rows, including video error rows. Video failures do not increment the run-level input error count. The OUTPUT record also reports delivered results, successful charge counts, completion information, and any warnings.

Errors and unavailable videos

An unparseable input produces a row with the original input, a null video_id, and an error string. It has no canonical video URL. A parseable but unavailable video keeps its id and canonical URL and includes an error such as VideoUnavailable. These two rows come from an actual run:

[
  {"input": "not a youtube link", "video_id": null, "fetched_at": "2026-10-02T02:27:49.089Z", "error": "could not parse video id"},
  {"input": "zzzzzzzzzzz", "video_id": "zzzzzzzzzzz", "fetched_at": "2026-10-02T02:27:49.089Z", "video_url": "https://www.youtube.com/watch?v=zzzzzzzzzzz", "error": "VideoUnavailable: The video is no longer available"}
]

Disabled subtitles, restricted videos, and temporary video access failures also appear as individual error rows.

One failed video does not discard the successful transcripts from other videos. A run containing only video errors still completes and returns those error rows. No video event is charged for them. Input errors are recorded separately in the key-value store, using the established bad_input code.

Pricing

The price is $0.005 per successfully delivered, nonempty video transcript. There is one chargeable event, video, and no transcript start fee. Video error rows and empty transcript text are free. Requesting segment timestamps does not add a separate event. Requesting the same successful video twice produces two transcript rows and two video events.

Users without a paid Apify plan receive at most ten results per run. Video error rows also occupy a result position, but remain free. Nine successful transcripts and one video error count as ten delivered rows and nine chargeable transcripts. Paying users can submit larger lists. Large lists are processed in batches automatically; inspect the completion information if a run reaches its time limit.

Use cases for YouTube transcripts

• AI and RAG pipelines. Feed the transcript text into a language model to summarize a talk, answer questions about a lecture, or build a searchable knowledge base from a channel's videos.
• Content repurposing. Turn a video into a blog post draft, show notes, a newsletter or social posts without typing it out.
• Research and monitoring. Collect what speakers, competitors or creators say across many videos and search the text for brands, products or topics.
• Subtitles and accessibility. Keep segments_data to align quotes with playback times, create chapter markers, or check a caption track.
• Language work. Request the same video in several languages through preferred_languages and compare the tracks the video offers.

YouTube transcript API: using the results

Download the dataset from the Apify console as JSON, CSV or Excel, or read it with the Apify API from your own code. The actor works as a YouTube transcript API: start a run with a list of links and fetch the dataset items when it finishes. Use transcript for text search, summarization, research notes, or document ingestion. Use segments_data when you need to connect a passage to a playback timestamp. Keep video_id, video_url, and language alongside the text so your application can identify the source and the selected track.

Comparison with other ways to get a YouTube transcript

Approach	Bulk lists	Language choice	Timestamps	Setup
Copying "Show transcript" in the YouTube player	No, one video at a time	Only what the player shows	Visible, not exportable	None, but manual
Running an open-source transcript library yourself	Yes, if you write the loop	Yes	Yes	Code, hosting and upkeep when YouTube changes
Speech-to-text transcription services	Yes	Depends on the model	Yes	Higher cost per minute of audio
This actor	Yes, one row per input in order	Ordered preferences with fallback	Optional per segment	Paste links and run

This actor reads the subtitles that already exist for a video, so results are fast and cheap and match what the video's captions say. When a video has no subtitles at all, a speech-to-text service is the better tool; this actor then returns a free error row instead of guessing.

FAQ

Can I request a specific language?

Yes. Put the preferred codes in preferred_languages, in priority order. The selected track must exist on the video. If your languages are unavailable, the actor falls back to an available track and reports its actual language.

Does this actor transcribe videos without captions?

No. It downloads existing subtitles. It does not perform speech recognition or produce estimated text when YouTube has no available subtitle track.

Are timestamps included by default?

No. Set include_segments to true for individual segment start times, durations, and text. Without that option, segments_data is absent rather than null.

Will one unavailable video fail my entire list?

No. Its row contains an error and is free. Other videos continue and successful transcripts retain the order of their inputs.

What does the free-plan limit count?

It counts delivered input positions, including video error rows, up to ten per run. Charges count only the nonempty transcripts among those delivered rows.

Is this YouTube transcript downloader free?

Every user can try it: users without a paid Apify plan get up to ten results per run. Each successful transcript costs $0.005; error rows and empty transcripts are always free.

Is this a YouTube transcript generator?

It downloads the transcript YouTube already has for a video, manual or auto-generated, so you get the same text the player shows under "Show transcript". It does not generate new text from the audio when a video has no subtitles.

Which links can I paste?

Standard watch links, youtu.be short links, mobile links, shorts, embed and /v/ links, and bare eleven-character video ids. Mixed lists are fine.

Can I export transcripts to CSV or Excel?

Yes. The dataset downloads as JSON, CSV, Excel and other formats from the Apify console or the API.