Twitter Transcript Scraper
The speech in an X (Twitter) post's video as text, in typed JSON: transcript, detected language, duration, status and video URL. Pick a model size, translate to English if you want. 2 credits per call. No X login.
Free API key · 2 credits per transcript · failed calls free
// GET /api/v2/twitter-transcript/?urls=https://x.com/MrBeast/status/2096285173788840001 · sample from the catalog { "success": true, "data": { "urls": "https://x.com/MrBeast/status/2096285173788840001", "transcript": { "input_url": "https://x.com/MrBeast/status/2096285173788840001", "transcript": "We're in the middle of the world's deadliest show...", "language": "en", "status_id": "2096285173788840001", "video_url": "https://video.twimg.com/....mp4", "duration": 31.2, "status": "success" }, "next_cursor": null } }
| field | value | type | Has a value✓ |
|---|
Scrapier's Twitter Transcript Scraper turns the video in an X post into text. Send the post URL and get JSON with transcript, language, duration, status, status_id and video_url. It costs 2 credits per successful call. No X login. The call can run long: use a 210-second timeout.
Twitter Transcript Scraper reads the video of an X post and returns what is said as text. You give it a post URL; it fetches the video, transcribes the audio with a faster-whisper model you can size from tiny to large-v3, and returns the text, the detected language, the duration and the video URL. You can force a language, translate the result to English or drop non-speech audio. Every response is structured JSON, ready to pass directly to an LLM, index into a vector store or feed a captioning pipeline. It is one of 6 X (Twitter) endpoints on the same API key used by 12,000+ developers.
What data can I extract with Twitter Transcript Scraper?
Using this Twitter Transcript Scraper, you can extract the following data from public X (Twitter) transcripts:
Want the exact field names? See the example output.
Try Twitter Transcript Scraper Live
Press Run to see the catalog sample. Add your API key and it runs a real call to the Scrapier API from your browser. The key is sent only to proxy.scrapier.io and never stored.
+ Run live with your API key
How to use Twitter Transcript Scraper
Hand it to your AI assistant, or run it yourself in five steps.
Copy to your AI assistant
Paste this into ChatGPT, Claude or Cursor and it knows how to call the scraper.
Use the Scrapier Twitter Transcript Scraper to get public X (Twitter) transcripts. Endpoint: GET https://proxy.scrapier.io/api/v2/twitter-transcript/?urls={urls} with header "X-API-Key: YOUR_API_KEY" and a 210-second timeout. Send one value per call in urls; for a list, make one call per value, at most 3 in parallel. It returns JSON under data.transcript: input_url, transcript, language, status_id, video_url, duration, status. Each successful call costs 2 credits; failed calls are free. If you support MCP, run: claude mcp add -t http scrapier https://mcp.scrapier.io/mcp --header "X-API-Key: YOUR_API_KEY" and call run_scrape with endpoint "twitter-transcript" (find it with search_endpoints).Use this scraper from ChatGPT, Claude or Cursor. Point your assistant at Scrapier and it reads X (Twitter) transcripts on its own: no glue code, no scraping logic in your prompts.
Run it yourself
- Get your free Scrapier API keySign up at app.scrapier.io; the free tier lets you try it first.
- Open Twitter Transcript Scraper in the catalogIt lives under Scrapers / X (Twitter).
- Enter the post URLPut the URL of an X post with a video in
urls, such ashttps://x.com/MrBeast/status/2096285173788840001. Optionally setmodel_size,languageandtranslate_to_english. - Send your API requestOne GET with your key in the
X-API-Keyheader, or onerun_scrapecall from your agent. This endpoint is long running: set a 210-second client timeout. - Save your resultsThe text comes back as JSON in
data.transcript.transcript, withlanguage,durationandstatusbeside it. Store it as is, or write it to a file.
How to scale to bulk X (Twitter) transcripts
urls takes one value per call, so for a list you loop: one request per post. There is no paging: data.next_cursor is always null. Run up to 3 requests in parallel per API key; above that the API answers 429 with a Retry-After header, and throttled calls cost nothing. Each call is 2 credits and failed calls are free, so 1,000 calls cost 2,000 credits, inside the Hobby plan's 15,000.
Features
Twitter Transcript Scraper turns X video into text you can search, quote and feed to a model. It suits media monitoring, accessibility, research, content repurposing and any workflow that needs what was said in a post's video, without downloading and transcribing it yourself.
- Transcript
transcriptholds the spoken text of the video - Language
languageis detected, or forced with thelanguageinput - Translation
translate_to_englishreturns English text - Model size
model_sizefromtinytolarge-v3 - Speech only
vad_filterdrops non-speech audio - Length cap
max_duration_seconds(default 900) fails longer videos as a record
The download and transcription run on Scrapier's side, so you send one request and get JSON back, at 2 credits per transcript and 0 credits for a failed call.
Use cases
- Media monitor transcribes videos posted by public figures and searches
transcriptfor brand names and keywords. - Researcher turns a set of post videos into text and codes the transcripts for a study of political or product messaging.
- Content team repurposes a creator's video post into a written summary or an article draft from
transcript. - Accessibility editor adds a text version of a video post for readers who cannot play audio.
- AI engineer gives an agent a tool that transcribes a post's video, then asks an LLM to summarise or translate the text.
Running these from an agent? Use the MCP integration and let Claude or Cursor call the scraper directly.
Why choose this Scraper API?
The transcript is produced from the audio of the post's video, so it works for videos that carry no captions.
model_size, beam_size, language and vad_filter let you trade run time for accuracy, and translate_to_english adds translation.
The same key and response envelope cover posts, profile tweets and video transcripts, so a post URL from one call feeds the next.
Input
The scraper takes one X post URL per call in urls. All other inputs are optional: model_size, beam_size, language, translate_to_english, vad_filter and max_duration_seconds. It is a lookup: next_cursor is always null. Send parameters in the query string (GET) or as a JSON body (POST).
| Name | Type | Required | Default | Description |
|---|---|---|---|---|
urls | string | Yes | — | The URL of an X post with a video. One per call. Required |
translate_to_english | boolean | No | false | true translates the transcript to English (default false) |
model_size | string | No | small | faster-whisper model size: tiny, base, small (default), medium or large-v3 |
beam_size | integer | No | 5 | Beam search width, 1 to 10 (default 5) |
language | string | No | — | ISO language code to force; empty (default) means auto-detect |
vad_filter | boolean | No | true | true (default) drops non-speech audio |
max_duration_seconds | integer | No | 900 | Videos longer than this fail as a record (default 900) |
use_proxy | boolean | No | true | Use Scrapier's residential proxy pool when true (default true). Leave as is |
raw_data | boolean | No | false | Return X's raw upstream payload under raw_responses instead of the formatted transcript |
One post per call: to transcribe ten posts, make ten calls. Booleans are the strings true or false. The endpoint is long running, so keep the 210-second client timeout.
Example input
{
"urls": "https://x.com/MrBeast/status/2096285173788840001",
"translate_to_english": "false",
"model_size": "small",
"beam_size": "5",
"language": "",
"vad_filter": "true",
"max_duration_seconds": "900",
"use_proxy": "true",
"raw_data": "false"
}Error codes
| Status | Meaning | Billed | Fix |
|---|---|---|---|
| 200 | Transcript data returned in data.transcript | Yes | — |
| 400 | Bad or missing input | No | Check urls: one value per call |
| 401 | X-API-Key header missing | No | Add the header |
| 402 | Out of credits (body has balance) | No | Top up; retrying will not help |
| 403 | Invalid or revoked API key | No | Check the key in your dashboard |
| 429 | Throttled | No | Wait Retry-After seconds, then retry |
| 5xx | Upstream failure or timeout | No | Retry with backoff; set a 210 s client timeout |
From the error reference. Any non-2xx response costs 0 credits.
Output
Typed JSON with the same schema on every call. One call returns one transcript object under transcript with 7 fields, inside the standard data envelope. data also carries your urls input and next_cursor. Field names stay the same on every call. Set raw_data to true if you want X's raw payload instead.
Example output
{
"success": true,
"data": {
"urls": "https://x.com/MrBeast/status/2096285173788840001",
"transcript": {
"input_url": "https://x.com/MrBeast/status/2096285173788840001",
"transcript": "We're in the middle of the world's deadliest show...",
"language": "en",
"status_id": "2096285173788840001",
"video_url": "https://video.twimg.com/....mp4",
"duration": 31.2,
"status": "success"
},
"next_cursor": null
}
}The catalog's sample response for urls=https://x.com/MrBeast/status/2096285173788840001, labelled as a sample: live calls return the account's current values.
Transcript
| Field | Type | Example | Description |
|---|---|---|---|
transcript.transcript | string | "We're in the middle of the world's deadlies…" | The spoken text of the video |
transcript.language | string | "en" | Language code of the speech (en in the sample) |
transcript.duration | number | 31.2 | Duration of the video (31.2 in the sample) |
transcript.status | string | "success" | Result of the transcription (success in the sample) |
Post and video
| Field | Type | Example | Description |
|---|---|---|---|
transcript.input_url | string | "https://x.com/MrBeast/status/20962851737888…" | The post URL you sent |
transcript.status_id | string | "2096285173788840001" | ID of the X post |
transcript.video_url | string | "https://video.twimg.com/....mp4" | URL of the video file |
Response envelope
| Field | Type | Example | Description |
|---|---|---|---|
success | boolean | true | true on every 200 response |
data.urls | string | "https://x.com/MrBeast/status/20962851737888…" | Your post URL, echoed back |
data.next_cursor | null | null | Always null: this is a lookup with no paging |
How does it work?
Scrapier sends each request through its residential proxy pool (use_proxy, on by default), so you never manage IPs or proxy accounts.
Retries and parsing of X's responses happen on Scrapier's side. You send one HTTP request and get typed JSON back.
Only publicly visible data is returned: what any visitor sees without signing in. The post must contain a video with speech.
The schema is stable. When X changes, Scrapier updates the scraper and the field names stay the same.
How do you integrate Twitter Transcript Scraper into your stack?
It works with what you already use: any HTTP client, AI agents over MCP, and no-code automation tools. Code samples for cURL, Python and Node.js are on the API tab.
MCP integration for AI agents
This registers the Scrapier MCP server, whose tools reach every Scrapier scraper, including twitter-transcript: the agent finds it with search_endpoints and calls it with run_scrape. Works with Claude Code, Cursor and Windsurf. MCP calls are billed like REST: 2 credits per successful call. Full MCP setup →
No-code tools (n8n, Make, LangChain)
Use the HTTP Request node with your Scrapier API key in an X-API-Key header and point it at /api/v2/twitter-transcript/ with the input as a query parameter. Map data into the next node.
Add an HTTP "Make a request" module with the same header and URL, and parse the response as JSON. Map the fields you need straight into Sheets, Airtable or your CRM.
Wrap the endpoint in a requests-based tool, or load the Scrapier MCP server through LangChain's MCP adapters so the agent finds twitter-transcript with search_endpoints itself.
Pricing and Rate Limits
One X (Twitter) transcript costs 2 credits, so 1,000 transcripts cost 2,000 credits: $3.16 on the Pro plan. Failed and empty calls are not billed. Run up to 3 requests in parallel per key; above that the API answers 429 with a Retry-After header.
| Plan | Monthly price | Credits | Transcripts included | Per 1,000 transcripts |
|---|---|---|---|---|
| Hobby | $29 | 15,000 | 7,500 | $3.87 |
| Pro | $79 | 50,000 | 25,000 | $3.16 |
| Scale | $299 | 250,000 | 125,000 | $2.39 |
| Enterprise | Custom volume and terms | |||
Prices from the live billing page. See all plans and limits.
Call cost and timing
How we measure this
Credit cost, pagination and the sample come from the live catalog (GET /api/v2/endpoints, /twitter-transcript/, checked 9 Nov 2026). Concurrency and timeout guidance come from the rate limits page of the docs. Scrapier does not publish latency or success-rate figures for this endpoint, so none are shown here.
Twitter Transcript Scraper vs the Official X (Twitter) API
| Feature | Scrapier transcript scraper | Official X (Twitter) API |
|---|---|---|
| Access needed | One Scrapier API key | A developer account, app credentials and a plan with access limits |
| Transcript | Text from the post's video, with language and duration | No transcription: you download and transcribe the media yourself |
| Fields returned | 7 fields: text, language, duration, status, post ID, video URL | Post data and media links, depending on your plan |
| Rate limits | Up to 3 parallel requests per key; 429 with Retry-After | Limits set per endpoint and plan |
| Setup | Sign up, copy a key, send a request | Apply for access, create an app, run a speech-to-text model |
| Cost | 2 credits per transcript, from $2.39 per 1,000 calls | Set by the plan you choose, plus your own transcription cost |
The official X API is the better choice when you post or act on behalf of a user, or need data that only the account owner can see.
Is it legal to scrape X (Twitter) data?
Scraping publicly available X (Twitter) data is generally lawful in many countries, but the answer depends on what you collect, where you are and how you use it. In the US, hiQ Labs v. LinkedIn is the case most often cited on public-data scraping. Scrapier returns only public video posts any visitor can see without logging in. Under GDPR and similar laws, collecting public data can be permissible, but how you store and use personal data, such as names, handles or posts, must still comply with local data-protection rules, and X's terms apply. Consult legal counsel if your use case involves bulk storage of personal data. This is not legal advice; see our Terms of Use.
Frequently asked questions
What fields does Twitter Transcript Scraper return?
The transcript object has transcript, language, duration, status, input_url, status_id and video_url: 7 fields in data.transcript. The data envelope adds urls and next_cursor. Each successful call costs 2 credits and needs no X login.
How do I transcribe an X video?
Send the post URL in urls and read data.transcript.transcript. Scrapier fetches the video, transcribes the audio and returns the text with the detected language and duration. Keep a 210-second client timeout, because the endpoint is long running.
Can it translate the transcript to English?
Yes. Set translate_to_english to true and the transcript is returned in English. Leave language empty to auto-detect the spoken language, or set an ISO language code to force it.
Which model sizes can I choose?
model_size accepts tiny, base, small (the default), medium and large-v3, which are faster-whisper models. beam_size sets the beam search width from 1 to 10, default 5. Smaller models are quicker; larger ones are generally more accurate.
What happens with a very long video?
Videos longer than max_duration_seconds, 900 by default, fail as a record instead of being transcribed. Raise the value if you need longer videos, and keep the 210-second client timeout. Failed calls are not billed.
Does the post need a video?
Yes. The endpoint transcribes the video of the post you send in urls. A post without a video or speech has nothing to transcribe, so check the status field and the HTTP status of the call before using the text.
Does Twitter Transcript Scraper require an X account or login?
No. Twitter Transcript Scraper uses no X account or cookies: you send your Scrapier API key in the X-API-Key header and the input, and get JSON back. Only public X data is returned, and no X developer account or API credentials are involved.
Does Twitter Transcript Scraper work with Claude, ChatGPT and other AI agent tools?
Yes. Connect the Scrapier MCP server with claude mcp add -t http scrapier https://mcp.scrapier.io/mcp and your API key in an X-API-Key header; it works in Claude Code, Cursor and Windsurf. The agent finds twitter-transcript with search_endpoints and calls it with run_scrape. Any agent that can call an HTTP endpoint can use the REST API directly.
What happens if the input is wrong or X returns nothing?
If a call fails, the API returns a non-2xx status and costs 0 credits. In code, check the status, then skip that input or retry it later. Throttled calls (429 with a Retry-After header) are free as well, so retry loops cost nothing.
Does it return data in a format LLMs can use directly?
Yes. Every response is typed JSON with the same field names on every call: no HTML parsing and no selectors. Pass it straight to an LLM, index it into a vector store or return it from an agent tool.
How much does it cost to run Twitter Transcript Scraper 1,000 times?
2,000 credits, because each call is 2 credits. That is about $3.87 on Hobby, $3.16 on Pro and $2.39 on Scale, at the plans' price per credit. Failed calls are free.
How do I use Twitter Transcript Scraper with Python?
Send a GET request with the requests library to https://proxy.scrapier.io/api/v2/twitter-transcript/ with params={"urls": "https://x.com/MrBeast/status/2096285173788840001"}, your key in the X-API-Key header and timeout=210. Read r.json()["data"]["transcript"]["transcript"]: it is already JSON, so no HTML parsing is needed.
Is it legal to scrape X (Twitter) data?
It depends on your use and your country, so this is not legal advice. Scrapier returns public X data only. X's terms and privacy law such as GDPR or CCPA apply to how you store and use it. Collect only what you need, and ask a lawyer about your specific case.
Your feedback
Found a bug or missing a field? Let us know. Write to support@scrapier.io and the team replies to every report.
Start Scraping X (Twitter) Transcripts
Public X (Twitter) transcripts as JSON, 2 credits each, from REST or MCP.