# What is said on YouTube, as text ready for your model. Canonical URL: https://www.rastriq.es/en/youtube-transcript-data/ · Spanish version: https://www.rastriq.es/datos-transcripciones-youtube/ > Transcripts from YouTube videos, Shorts, channels, playlists and searches as text, timestamped segments, SRT or WebVTT. No API key; pay per transcript. The captions of any YouTube video, Short or live stream, or of whole channels, playlists and searches, turned into clean text, timestamped segments, SRT or WebVTT, together with the video metadata. No API key and no login: you only pay for rows that return a transcript. QUICK ANSWER **Quick answer:** Transcripts from YouTube videos, Shorts, live streams, channels, playlists and searches as text, timestamped segments, SRT or WebVTT, with language, caption type, word count and video metadata; no API key and you only pay per transcript. Sample of 65 rows in CSV and JSON. $2 Per 1,000 transcripts (Free plan) 64 of 65 Videos with a transcript in the samples 24.8 h Of Spanish video turned into text in 31 s See use cases → See real rows CONTEXT ## Four kinds of input, one row per video Mix URLs, IDs, channels, playlists and searches in the same run. You choose the output formats: text, segments, timestamped text, SRT and WebVTT. | Input | Example | What you get | |---|---|---| | Video, Short or live stream | `https://www.youtube.com/watch?v=…` · `youtu.be/…` · bare ID | One row with its transcript | | Channel | `https://www.youtube.com/@TED` (add `/shorts` or `/streams`) | The latest videos, with a limit and a date filter | | Playlist | `https://www.youtube.com/playlist?list=…` | The playlist videos up to your limit | | Search | `"searchQueries": ["what is an AI agent"]` | The top videos for each query | ### There is a row even when there are no captions | Status | Meaning | Charged | |---|---|---| | OK | Transcript returned | Yes | | NO_TRANSCRIPT | The video has no captions | No | | LANGUAGE_NOT_AVAILABLE | None of your languages exists and fallback is off | No | | UNAVAILABLE | Private, removed, age-restricted or geo-blocked | No | | BLOCKED · ERROR | YouTube blocked the request or an unexpected error occurred | No | | INVALID_INPUT | The input is not a recognisable YouTube URL, ID or query | No | **This is not speech-to-text.** The actor reads the captions YouTube already has, uploaded by the creator or auto-generated, and does not translate. If a video has none you get `NO_TRANSCRIPT` at no cost: in the TED sample the only case is a 90-second music performance. USE CASES ## What YouTube transcripts are used for Six real use cases: who uses them, what you do and which fields you need. 1 · Knowledge base for RAG AI and product teamsLoad every video of a channel (training, webinars, conferences) into your vector store. Timestamped text (`[00:01:23] …`) lets the assistant cite the exact minute.**Result:** a per-channel corpus ready to chunk and embed, linking back to each moment of the video. Fields: `timestampedText`, `segments`, `videoId`, `url` 2 · Content research Agencies, marketing teams and analystsTranscribe the top videos for a query ("AI for business") and use an LLM to analyse which topics, arguments and keywords keep coming up.**Result:** a map of what is being said about a topic. Fields: `inputSource`, `title`, `text`, `viewCount`, `publishedAt` 3 · Weekly channel monitor Brand, communications and product teamsSchedule your competitors' channels with `publishedAfter: "7 days"` and receive only the new videos with their transcript: launches, messages and prices said out loud.**Result:** a weekly digest of what they published. Fields: `channelName`, `publishedAt`, `title`, `text` 4 · Repurposing and SEO Content teamsTurn your own videos into articles, summaries, chapters or FAQs. `wordCount` and `durationSeconds` help size the work.**Result:** drafts from what is already recorded. Fields: `text`, `wordCount`, `durationSeconds`, `keywords` 5 · Subtitles and accessibility Media, training and eventsExport ready-to-use SRT or WebVTT files, in your preferred language among those available.**Result:** subtitle files per video. Fields: `srt`, `vtt`, `availableLanguages`, `transcriptLanguage` 6 · Evaluation and training corpora ML and research teamsCollect transcripts by language, length or topic, and separate creator captions from auto-generated ones with `isAutoGenerated`.**Result:** a corpus labelled by language and caption quality. Fields: `isAutoGenerated`, `transcriptLanguage`, `originalLanguage`, `wordCount` WHAT THE DATA SAYS ## Two real runs, in numbers What it costs, and how long it takes, to turn hours of video into text. | Task | Videos | With transcript | Words | Video hours | Auto-generated captions | Run time | Cost | |---|---|---|---|---|---|---|---| | TED channel, last 30 days | 25 | 24 | 48,586 | 5.4 | 3 of 24 | 19 s | $0.05 | | Two Spanish searches | 40 | 40 | 244,885 | 24.8 | 36 of 40 | 31 s | $0.08 | Runs of the two saved tasks on 16 October 2026, with 256 MB of memory. TED uploads its own captions in several languages: one talk has 14 tracks. In the Spanish videos found by search almost everything is auto-generated. That is why `isAutoGenerated` matters: it tells you what quality of text you are working with. As a rule of thumb, a TED talk of about 12 minutes is some 1,600 words, and an hour of Spanish video some 10,000. REAL SAMPLE ## Real rows from two saved tasks Extracts from runs on 16 October 2026. Public samples download without signing up. ### TED talks from the last 30 days | title | durationSeconds | viewCount | wordCount | languages | isAutoGenerated | |---|---|---|---|---|---| | How to Start Any Conversation the Right Way \| Chris Fenning \|… | 667 | 209366 | 1759 | 5 | false | | Why We Blew Up Four Dams — and Saved a River \| Amy Bowers… | 777 | 81041 | 1477 | 4 | false | | The Three Words That Saved My Life \| Blake Mycoskie \| TED | 937 | 120282 | 1703 | 5 | false | | Why I’m Building Libraries Inside Prisons \| Reginald Dwayne… | 749 | 46290 | 1644 | 9 | false | | How to Clean Up the World’s Dirtiest Industry \| Joojin Kim \|… | 459 | 99854 | 999 | 6 | false | | Why the Future is Human … Plus AI \| Esther Perel and Sal Khan | 1247 | 99646 | 3592 | 4 | false | Real rows from the "TED talks from the last 30 days" task. `languages` is the number of caption tracks in `availableLanguages`. ### Spanish videos from two searches | title | durationSeconds | wordCount | transcriptLanguage | isAutoGenerated | |---|---|---|---|---| | 10 Herramientas de Inteligencia Artificial para NEGOCIOS \| 🤖… | 2097 | 5926 | es | true | | ¿Se puede crear una empresa solo con IA? | 909 | 2879 | es | true | | GUÍA COMPLETA: INTELIGENCIA ARTIFICIAL para EMPRENDEDORES \|… | 7650 | 23850 | es | true | | Como implementar Inteligencia artificial en mi empresa 2026 | 874 | 2744 | es | true | | Los 5 Mejores Negocios Con Inteligencia Artificial (Probados) | 857 | 2378 | es | true | | Cursos de IA para Negocios 2026 \| Cursos de Inteligencia… | 12057 | 26351 | es | true | Real rows from the "Spanish searches" task. The channel is left out of these rows because many are individual creators; the actor does return it. ### The timestamped format (`timestampedText`) ``` [00:00:04] I want to be long-winded and unclear. [00:00:11] It's not a common opening line. [00:00:13] Most people won't tell us when they're going to be long-winded, ``` **Only an excerpt.** The samples carry the first words of each transcript in `text_excerpt`. The full text belongs to its authors and you get it by running the actor. [YouTube transcripts CSV](https://www.rastriq.es/muestras/youtube-transcripciones.csv) [YouTube transcripts JSON](https://www.rastriq.es/muestras/youtube-transcripciones.json) [Browse all samples](https://www.rastriq.es/en/samples/) RECIPES ## Inputs ready to copy Paste the input into the actor's JSON tab and press Start. The first two are the saved tasks used to produce the samples. | Task | Input (JSON) | What it is for | |---|---|---| | TED channel, last 30 days | `{"urls": ["https://www.youtube.com/@TED"], "publishedAfter": "30 days", "languages": ["en"], "outputFormats": ["text", "timestamped"], "maxVideosPerSource": 40}` | A recent channel corpus, with timestamps to cite. | | Spanish searches | `{"searchQueries": ["inteligencia artificial para empresas", "qué es un agente de IA"], "languages": ["es"], "outputFormats": ["text", "segments"], "maxVideosPerSource": 20}` | Researching a topic on video. | | One video as plain text | `{"urls": ["https://www.youtube.com/watch?v=jNQXAC9IVRw"], "outputFormats": ["text"]}` | The quickest test. | | A channel's Shorts | `{"urls": ["https://www.youtube.com/@TED/shorts"], "maxVideosPerSource": 50}` | Short videos with their text. | | SRT and WebVTT subtitles | `{"urls": ["https://www.youtube.com/playlist?list=…"], "languages": ["es", "en"], "outputFormats": ["srt", "vtt"]}` | Subtitle files for a whole playlist. | | Weekly channel monitor | `{"urls": ["https://www.youtube.com/@canal-1", "https://www.youtube.com/@canal-2"], "publishedAfter": "7 days", "outputFormats": ["text"]}` | Schedule it every Monday: new videos only. | Control volume with `maxVideosPerSource` (per channel, playlist or search) and `maxVideos` (whole run). Languages go in order of preference: the first available one wins. FIELDS ## Field dictionary One row per processed video; text formats appear according to the ones you choose. | Field | Description | |---|---| | `videoId · url · title · inputSource` | Video and the input it came from (URL, channel or search) | | `channelName · channelId · channelUrl` | Channel | | `durationSeconds · viewCount · likeCount · publishedAt` | Duration, views, likes and publish date | | `isLive · isLiveContent` | Live right now / live or past stream | | `transcriptStatus` | OK, NO_TRANSCRIPT, LANGUAGE_NOT_AVAILABLE, UNAVAILABLE, BLOCKED, ERROR, INVALID_INPUT | | `transcriptLanguage · originalLanguage · isAutoGenerated` | Returned language, original language and whether captions are auto-generated | | `availableLanguages` | Every caption track with code, name and whether it is auto-generated | | `segmentCount · wordCount` | Number of segments and words | | `text` | Transcript as clean text | | `segments` | List of segments with start, duration, end and text | | `timestampedText` | Text with an [hh:mm:ss] mark per line | | `srt · vtt` | Subtitles in SRT and WebVTT format | | `description · keywords · thumbnailUrl` | Optional metadata (on by default) | DELIVERY ## How you receive the data Run it yourself, receive it on a schedule or commission the project. Apify You run the actor and download JSON, CSV or Excel. $2 per 1,000 transcripts on the Free plan, $1.80 on Starter, $1.50 on Scale and $1.20 on Business. No start fee and proxy included; rows without a transcript are free. Channels: console, API, MCP Recurring dataset Your channels or searches every week, with only the new videos, in the format your system needs. Channels: S3, CSV, JSON Custom project Chunking, embeddings and loading into your vector store, per-video summaries and topic classification. [Rastriq for AI →](https://www.rastriq.es/en/ai/) FAQ ## Frequently asked questions about YouTube transcripts ### Does it transcribe the audio? No. It returns the captions YouTube already has, from the creator or auto-generated. Videos without captions give a `NO_TRANSCRIPT` row that is not charged. ### Does it work with Shorts, live streams and playlists? Yes. It accepts Shorts and live URLs, channels with the `/shorts` or `/streams` tab, playlists and searches, all mixed in one run. ### Which language does the text come in? In the first language of your `languages` list that exists. If none does, you get the default track (or `LANGUAGE_NOT_AVAILABLE` if you turn fallback off). There is no machine translation; `availableLanguages` lists every track. ### What about age-restricted videos? They require a login, which the actor does not use, so they return `UNAVAILABLE` at no cost, like private, removed or geo-blocked videos. ### Can I use it from an AI agent? Yes. Call it through the Apify API or the Apify MCP server; the timestamped text is designed for prompts and for citing the exact moment of the video. ### Is it legal to use these transcripts? The actor only reads public captions. You are responsible for complying with YouTube's terms, copyright and data protection when using the output. Rastriq is not affiliated with YouTube or Google. FREE SAMPLE ## Get an extract for your channels or searches Tell us which channels, playlists or topics you care about and we will send an extract in the schema you would receive. - Videos, Shorts, channels, playlists and searches - Text, segments, SRT and WebVTT - You only pay for what has a transcript Received. Meanwhile, download the real samples at [/en/samples/](https://www.rastriq.es/en/samples/). RELATED PAGES ## Keep exploring [Rastriq for AI Data and files designed for agents.](https://www.rastriq.es/en/ai/) [Competitive intelligence Traffic, trends and video.](https://www.rastriq.es/en/competitive-intelligence/) [TikTok data Ads, trends and creators.](https://www.rastriq.es/en/tiktok/) [Buy datasets Catalogue with downloadable samples.](https://www.rastriq.es/en/buy-datasets/)