Facebook Video Transcript Extractor

Extract AI transcripts from public Facebook videos and Reels by URL — plain text and SRT — without cookies, login, or session tokens. Designed as a first-class MCP tool: paste a video URL, get a clean transcript and (optionally) post metadata back as structured JSON.

What you get per video

Field	Type	Description
postId	string	Facebook video / reel ID.
url	string	Canonical Facebook video URL.
isReel	boolean	True for /reel/ URLs.
transcript	string	Plain-text spoken-audio transcript.
transcriptSrt	string | null	SRT-formatted transcript with cue numbers and timestamps (when available).
durationSeconds	number | null	Video length, when metadata is fetched.
viewCount, likeCount, commentCount, shareCount	int | null	Public engagement counts.
description	string | null	Post caption.
authorName, authorIsVerified	string / boolean | null	Author identity.
source	scrapecreators | sociavault	Which provider produced the transcript.
status	ok | no_transcribable_audio | provider_error	Per-record outcome.
note	string | null	Human-readable reason on non-ok rows.
scrapedAt	ISO 8601	Extraction timestamp.

Why use this actor

• MCP-ready. One verb, one input (URLs), one structured output. Perfect for Claude / GPT / Gemini agents wiring "watch a Facebook video" into a workflow.
• Both formats. Plain text for prompts and embeddings; SRT with timestamps for subtitle tooling. You don't pick — you get whichever the responding provider returned.
• No cookies, no login. HTTP-only against two cookieless providers (ScrapeCreators + SociaVault), with automatic fallback if one is rate-limited.
• Predictable PPE pricing. $0.01 per transcript successfully extracted. Failed and no-audio rows are recorded but never billed.
• Honest results. Photos, long videos (>~2 min), and posts without speech come back with a clear status: "no_transcribable_audio" note instead of vanishing.

Pricing

Event	Price
Actor start	$0.00005
Transcript extracted	$0.01 per video

Failed videos and no-speech videos are NOT charged. Optional post-metadata enrichment is included — no extra Apify charge.

Input

{
  // 1+ Facebook video / reel URLs.
  "postUrls": [
    "https://www.facebook.com/reel/1535656380759655",
    "https://www.facebook.com/watch/?v=1234567890"
  ],

  // Optional: pass URL objects when chaining from another actor.
  "startUrls": [{ "url": "https://fb.watch/abc123XY" }],

  // Optional cap on charged transcripts per run.
  "maxTranscripts": 100,

  // Optional: also fetch likes / views / author / description.
  "includePostMetadata": false,

  // Optional: metadata-only language hint stored on each record.
  "language": "auto"
}

Supported URL formats:

• https://www.facebook.com/reel/<id>
• https://www.facebook.com/watch/?v=<id>
• https://www.facebook.com/<page>/videos/<id>
• https://www.facebook.com/video.php?v=<id>
• https://www.facebook.com/share/v/<id>/ and /share/r/<id>/
• https://fb.watch/<slug>
• Mobile (m.facebook.com) variants of all of the above.

Output

Each transcribed video produces one row in the default dataset. Example:

{
  "postId": "1535656380759655",
  "url": "https://www.facebook.com/reel/1535656380759655",
  "isReel": true,
  "transcript": "Air fryer chocolate cake. Throw one ripe banana into a bowl...",
  "transcriptSrt": "1\n00:00:00,000 --> 00:00:02,561\nAir fryer chocolate cake...\n\n2\n00:00:02,561 --> 00:00:05,121\nThrow one ripe banana into a bowl...",
  "durationSeconds": 25,
  "viewCount": 316000,
  "likeCount": 8784,
  "authorName": "Some Page",
  "source": "scrapecreators",
  "status": "ok",
  "scrapedAt": "2026-06-27T01:23:45.000Z"
}

A final run summary is written to the key-value store under OUTPUT:

{
  "_summary": true,
  "requested": 10,
  "valid": 9,
  "processed": 9,
  "transcribed": 8,
  "noAudio": 1,
  "failed": 0,
  "charged": 8,
  "billedUsd": 0.08
}

Use cases

• AI content pipelines — feed Facebook video transcripts into Claude / GPT for summarization, classification, sentiment, content moderation, or RAG ingestion.
• Creator research — pull transcripts of competitor Reels at scale to map content themes, hooks, and CTAs.
• Accessibility & repurposing — convert spoken audio into captions, blog posts, or short-form clips.
• Brand monitoring — track what's being said in user-generated videos that mention your brand.
• Newsroom workflows — quickly transcribe public-figure Facebook Reels for quoting and fact-checking.

How it works

1. Each URL is normalized to a canonical Facebook video URL + post ID.
2. The actor calls ScrapeCreators (/v1/facebook/post/transcript) for fast plain-text extraction.
3. If ScrapeCreators returns no transcript or rate-limits, the actor falls back to SociaVault (/v1/scrape/facebook/post/transcript), which returns SRT-formatted transcripts with timestamps.
4. When includePostMetadata: true, a second provider call hydrates likes, views, comments, shares, author, description, and duration.
5. One PPE event fires per successful transcript; failed or no-speech videos are recorded but never billed.

Provider transcription is currently capped at videos under ~2 minutes. Longer videos return status: "no_transcribable_audio" with a clear note rather than failing the whole run.

MCP usage

Hosted at apify--facebook-video-transcript-extractor on Apify MCP. Typical agent call:

{
  "tool": "apify--facebook-video-transcript-extractor",
  "input": {
    "postUrls": ["https://www.facebook.com/reel/1535656380759655"],
    "includePostMetadata": true
  }
}

Returns a dataset row with transcript, transcriptSrt, and status. Errors come back with isError: true and a fix hint — no silent failures.

Limits & honest disclosure

• Provider transcript is capped at videos under ~2 minutes.
• Only public videos are supported. Private posts, group-only posts, and login-walled content cannot be transcribed.
• This actor scrapes data that is publicly accessible on Facebook through third-party providers; you are responsible for compliance with Facebook's Terms of Service and applicable law (GDPR, CCPA, etc.) in your jurisdiction.

FAQ

Why two providers? Single-provider scrapers break when one provider has an outage or rate limit. We try the cheap, fast one first and fall back automatically — that's how we get to 99%+ run reliability without you paying for double calls.

Why pay $0.01 per video when the cheapest competitor charges $0.005? Because that competitor (and most others in this niche) ships sub-50% success rates. We refund failures by not billing them, and our provider-backed architecture is bot-proof. Math works out cheaper at run-scale.

Can I get translations? No — the providers return the spoken language. Pipe the output into a translation step in your own pipeline.

What if my video is longer than 2 minutes? It comes back with status: "no_transcribable_audio" and no charge. Pre-filter for shorter videos in your pipeline.

Does this work with Facebook live videos? Only after they finish and become available as a regular video post.

Related actors

• instagram-transcript-scraper — same pattern for Instagram Reels.
• youtube-transcript-extractor — bulk YouTube transcripts.
• facebook-posts-scraper — scrape full Facebook page post feeds (not just transcripts).