YouTube Transcript API & Scraper
Pricing
from $2.00 / 1,000 transcript delivereds
YouTube Transcript API & Scraper
Extract transcripts from YouTube videos, Shorts, playlists, and channels—no YouTube API key. Get clean text, timestamps, SRT, VTT, Markdown, metadata, and RAG-ready chunks for AI agents and automation. Original-language output; failed videos are free.
Pricing
from $2.00 / 1,000 transcript delivereds
Rating
0.0
(0)
Developer
PANIX AI LAB
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
0
Monthly active users
3 days ago
Last modified
Categories
Features
Share
Get clean, timestamped transcripts and subtitles from YouTube videos, Shorts, playlists and whole channels, ready for AI, RAG, research or repurposing. No YouTube API key.
- In the original spoken language by default, not a silent English translation or dubbed track.
- AI-ready formats: plain text, timed segments, SRT, WebVTT, Markdown and chapter-aware chunks for vector databases.
- One price per delivered transcript. Invalid links, unavailable videos and videos without captions are free.
Try it now: the input is already filled in (one video, with timed segments and video details). Click Start and the results appear in a table you can export to Excel, CSV or JSON. That example run costs under 1 cent, which the $5 of free monthly credit on Apify's free plan covers.
Price: from $2.00 per 1,000 transcripts, and you pay only for results you get. The full price table is in the Pricing section below.
Example result
One result, shortened. Every field is described in the Output section below.
{"videoId": "jNQXAC9IVRw","url": "https://www.youtube.com/watch?v=jNQXAC9IVRw","status": "ok","title": "Me at the zoo","channelName": "jawed","language": "en","isAutoGenerated": false,"transcript": "All right, so here we are in front of the elephants...","segments": [{"start": 0.0,"end": 3.0,"duration": 3.0,"text": "All right, so here we are..."}]}
What you get
- Videos, Shorts, playlists, and channels in one input.
- Original spoken language by default, instead of silently returning an English translation or dubbed track.
- One simple price per delivered transcript. Invalid URLs, unavailable videos, source errors, and videos without captions are free.
- Built for automation: stable video IDs, source provenance, deterministic ordering, and dedupe keys.
- AI-ready output: clean text, timestamps, Markdown, chapters, and vector-database chunks.
- No browser, video download, YouTube API key, or external AI service.
Use it from Apify Console, the API, MCP agents, n8n, Make, Zapier, LangChain, LlamaIndex, or a scheduled Actor task.
Quick start
- Paste a video, playlist, or channel URL.
- Keep Latest videos for recurring channel monitoring, or select All available for a historical import.
- Optionally enable SRT, VTT, Markdown, or RAG chunks.
- Click Start. Sort results by
positionto restore source order.
The prefilled example returns one public video transcript with timestamps and metadata.
{"videoUrls": ["https://www.youtube.com/watch?v=jNQXAC9IVRw"],"includeSegments": true,"includeMetadata": true}
Supported sources
| Input | Examples | Behavior |
|---|---|---|
videoUrls | Watch URLs, youtu.be, Shorts, live replays, embeds, mobile URLs, bare video IDs | Direct videos are processed first and do not consume maxVideos |
playlistUrls | Public playlist URL or playlist ID | Videos retain playlist order |
channelUrls | @handle, /channel/, /c/, /user/, /videos, /shorts, /streams | Bare channels include recent videos and Shorts |
You can mix all three. Videos are deduplicated globally by videoId; the first occurrence controls position, while every origin is retained in sources[].
Channel monitoring
Use channelMode: "latest" in a scheduled task to read up to 15 of the newest public uploads and Shorts from YouTube's compact channel feed. Use publishedAfter and publishedBefore for an inclusive UTC date window.
{"channelUrls": ["https://www.youtube.com/@TED"],"channelMode": "latest","publishedAfter": "2026-10-01","maxVideos": 25,"includeMarkdown": true,"includeChunks": true}
channelMode: "all" follows older-page continuations until maxVideos, the date boundary, or the end of the source.
Output
Each unique video produces one terminal dataset row:
recordType: "transcript"for a delivered transcript.recordType: "video_error"for a free per-video error.recordType: "source_error"when a playlist or channel cannot be expanded; also free.
Example transcript row, shortened:
{"schemaVersion": "2.0","recordType": "transcript","dedupeKey": "youtube:video:jNQXAC9IVRw","position": 0,"sourceType": "direct","videoId": "jNQXAC9IVRw","url": "https://www.youtube.com/watch?v=jNQXAC9IVRw","status": "ok","title": "Me at the zoo","channelName": "jawed","language": "en","isAutoGenerated": false,"transcript": "All right, so here we are in front of the elephants...","segments": [{"start": 0.0, "end": 3.0, "duration": 3.0, "text": "All right, so here we are..."}],"chargedEvents": {"transcript": 1},"charged": true}
Important fields:
| Field | Purpose |
|---|---|
position | Deterministic order across direct, playlist, and channel sources |
dedupeKey | Stable key (youtube:video:<videoId>) for recurring workflows |
sourceType, sourceUrl, sources[] | Where the video was discovered, including duplicate origins |
playlistId, playlistIndex | Playlist membership and source order |
channelId, channelUrl, publishedDate | Channel provenance and publication date. Dates are filled for channel runs (publishedDatePrecision: exact, or relative when derived from a label like "2 days ago") and are usually null for direct video and playlist inputs |
transcript, transcriptOriginal | Clean transcript in the original spoken language |
transcriptCaption | Cleaned text from YouTube's selected caption track |
segments | Timestamped caption lines |
srt, vtt, markdown, chunks | Optional output formats |
availableLanguages | Caption tracks YouTube exposes for the video |
warnings, warningCodes | Non-fatal quality or compatibility information |
errorCode, errorMessage | Classified free failure information |
A SUMMARY record in the run's key-value store counts discovered sources, unique videos, duplicates, date-filtered candidates, successful transcripts, failures, truncation, and spending-limit skips.
Input reference
| Field | Default | Description |
|---|---|---|
videoUrls | one prefilled example | Direct videos and Shorts |
playlistUrls | empty | Public playlists |
channelUrls | empty | Public channels or explicit channel tabs |
maxVideos | 50 | Unique videos discovered from playlists/channels, 1–1,000 |
channelMode | latest | Compact feed with up to 15 recent uploads/Shorts, or continuation-based all |
publishedAfter / publishedBefore | empty | Inclusive YYYY-MM-DD channel date filters |
language | empty | Empty selects the original spoken language |
preferManualCaptions | true | Prefer creator captions over auto-generated captions |
fixUrduScript | true | Correct Urdu written by YouTube's Hindi recognizer |
removeNonSpeechTags | false | Remove music, applause, and similar tags |
includeSegments | true | Timestamped segments |
includeMetadata | true | Video metadata and chapters |
includeSrt / includeVtt | false | Subtitle strings |
includeMarkdown | false | Timestamped LLM-ready Markdown |
includeChunks | false | Chapter-aware RAG chunks and deep links |
chunkSizeWords | 300 | Target chunk size |
chunkOverlapWords | 30 | Context overlap |
maxConcurrency | 10 | Parallel transcript requests |
maxRetries | 3 | Extra fresh-IP attempts on transient blocks |
At least one video, playlist, or channel source is required.
Pricing
There are only two events:
| Event | When it is charged |
|---|---|
apify-actor-start | Once per GB of configured memory, with a one-event minimum. Planned price: $0.003 |
transcript | Once after a usable transcript row is delivered |
All selected formats, metadata, playlist discovery, and channel discovery are included in the transcript price.
Planned transcript tiers:
| Tier | Per transcript | Per 1,000 |
|---|---|---|
| FREE | $0.00300 | $3.00 |
| BRONZE | $0.00275 | $2.75 |
| SILVER | $0.00250 | $2.50 |
| GOLD+ | $0.00200 | $2.00 |
The start event is $0.003. This covers lightweight scheduled channel checks even when no new transcript is found. The current effective prices are always shown on the Actor's Pricing tab.
You are not charged a transcript event for invalid URLs, duplicates, date-filtered candidates, unavailable/private/age-restricted videos, videos without usable captions, discovery errors, or retries. The Actor respects your maximum charge per run and stops dequeuing new videos when the limit is reached.
API examples
Python
from apify_client import ApifyClientclient = ApifyClient("<YOUR_APIFY_TOKEN>")run = client.actor("panix/youtube-transcript-scraper").call(run_input={"playlistUrls": ["https://www.youtube.com/playlist?list=YOUR_PLAYLIST_ID"],"maxVideos": 50,"includeMarkdown": True,"includeChunks": True,})for item in client.dataset(run.default_dataset_id).iterate_items():if item["recordType"] == "transcript":print(item["title"], item["transcript"][:200])
cURL
curl -X POST "https://api.apify.com/v2/acts/panix~youtube-transcript-scraper/run-sync-get-dataset-items" \-H "Authorization: Bearer <YOUR_APIFY_TOKEN>" \-H "Content-Type: application/json" \-d '{"videoUrls":["https://youtu.be/jNQXAC9IVRw"],"includeSrt":true}'
AI, RAG, and automation
- MCP and AI agents: expose the Actor through the Apify MCP server and return structured transcript records.
- RAG: enable
includeChunks; each chunk has a timestamp, word count, chapter, and deep link to the exact video moment. - n8n, Make, and Zapier: run the Actor, filter
recordType = transcript, and maptranscript,markdown, orchunksinto the next node. - Scheduled channel monitoring: save a task with
channelMode: latest, schedule it, and deduplicate downstream withdedupeKey.
An importable n8n workflow and a detailed channel-to-RAG recipe are included with the Actor source under examples/.
Reliability and language accuracy
Measured on the live Actor (October 2026):
- 20 languages, 20/20 transcripts, every one in the language actually spoken, in one run of about 30 seconds: English, German, French, Spanish, Hindi, Arabic, Urdu, Japanese, Italian, Filipino, Polish, Romanian, Dutch, Persian, Portuguese, Bengali, Tamil, Telugu, Ukrainian and Marathi. Where a video also had an English auto-generated track (Telugu, Ukrainian), the Actor still returned the spoken language.
- Hard cases: 17/17 eligible videos returned transcripts, including a 4-hour podcast, a 2.4-hour live replay, Shorts, auto-dubbed videos and unusual URL formats. Age-restricted, caption-less, deleted and invalid inputs each got a free, clearly labelled error row.
- Scale: 50/50 videos in about 45 seconds.
- Channels and playlists: a 40-video channel crawl returned 38 transcripts plus 2 free errors (no captions or no speech) in about 30 seconds; a playlist returned every video in playlist order.
- It uses direct mobile-player requests and JSON3 captions, without a browser or video/audio download.
About 1 in 5 regional news clips we sampled had no captions at all on YouTube. Those return NO_CAPTIONS and are never charged.
The Actor selects the original audio's caption track on auto-dubbed videos, prefers human-made captions by default, decodes HTML entities, removes rolling overlap, preserves Unicode, joins Chinese and Japanese correctly, and formats Arabic, Hebrew, Persian, Pashto, and Urdu subtitles for right-to-left display.
Urdu correction remains available as an advanced accuracy feature: if YouTube writes Urdu speech in Hindi/Devanagari script, transcriptOriginal is converted to Urdu script and transcriptCaption preserves YouTube's text.
English translation migration
Version 2 is a lightweight transcript product and no longer runs offline machine translation. Existing saved tasks remain safe:
- Legacy
outputLevel: originalstill returns text-only output. - Legacy
outputLevel: captionsstill returns timed output. - Legacy
outputLevel: englishdoes not fail or create a translation charge. - English videos keep the compatibility
transcriptEnglishalias. - Creator-uploaded English subtitles may still be returned.
- Other languages return the original transcript,
translationStatus: "not_available_in_core", and the warning codeENGLISH_TRANSLATION_MOVED.
Limitations
- The Actor extracts existing human-made or auto-generated captions. It does not run speech-to-text when a video has no captions.
- Private, members-only, unfinished live, and age-restricted videos are not supported.
- YouTube can change its internal player and browse responses. Automated canaries are used to detect breakage.
- Channel and playlist discovery covers public YouTube pages; malformed, private, or region-restricted sources return free source errors.
- Publication-date filters apply to channel candidates when a date is available.
publishedDatePrecision: "relative"identifies dates derived from labels such as “2 days ago”; unknown dates are retained rather than silently discarded. - You are responsible for complying with YouTube's Terms of Service, copyright rules, and applicable privacy laws.
FAQ
Do I need a YouTube API key?
No. The Actor uses public YouTube player and browse data.
Are failed videos charged?
No transcript event is charged unless a usable transcript row is delivered. The small Actor start event still applies to a run.
Can one bad video stop a playlist?
No. Every video is isolated and receives its own transcript or classified error row.
How do I avoid duplicates in scheduled channel runs?
Use dedupeKey, which is stable for the same YouTube video across runs.
Why might output arrive out of order?
Videos run concurrently. Sort by position to restore deterministic direct, playlist, and channel order.
Can it translate every transcript into English?
Not in this lightweight core version. It preserves original-language accuracy and existing creator-uploaded English subtitles. Offline machine translation may return later as a separate product if demand supports it.
Found a failing public source?
Open an issue in the Actor's Issues tab with the source URL and run ID. Never include private cookies or credentials.