Facebook Posts Media & OCR Content Harvester

Extract the visual content of public Facebook Page posts: a clean per-post image-asset manifest (full-resolution URL, width, height, id, suggested filename), the OCR image-text captured from Facebook's accessibility captions, and the entities found in the text — links, emails, phone numbers, mentions, hashtags, prices, and call-to-action phrases.

This actor is a superset sibling of a standard Facebook Posts Scraper. Where a posts scraper measures a timeline, this actor harvests the creative the scrape already fetched — with zero extra Facebook requests. Every emitted value is a copy or transform of something actually scraped (image URI / dimensions / id, accessibility caption, post text). Nothing is modeled or fabricated.

What you get

• 🖼️ Per-post image manifest — creative.images[] with url, width, height, id, ocrText, suggestedFilename, straight from media[].photo_image.
• 📄 OCR image-text — creative.imageText + hasTextInImage, derived from Facebook's accessibility caption (media[].accessibility_caption). Surfaced by no standard Posts scraper.
• 🔍 Entity extraction — creative.entities with real substrings of the text: urls, emails, phones, mentions, hashtags, prices, ctaMatches. Only non-empty types are attached.
• 🏷️ Derived type — photo / video / link / text, inferred from the scraped media typenames and entities.
• 📦 Run-level media manifest — a flat list of every image asset saved to the key-value store (MEDIA_MANIFEST) for bulk download, plus a MEDIA_INDEX summary (top OCR keywords, link domains, coverage counts).
• 🤖 Optional AI enrichment — per-post sentiment, topics, brandSafety, and language over the creative free-text (caption + OCR), across multiple providers.

Output modes

Input	Behaviour
default	Each dataset row = the pristine base post fields plus a namespaced creative block and surfaced view fields (type, imageCount, imageText, entitySummary, scrapedAt).
metricsOnly: true	Rows keep only identity fields + the creative block (drops the heavy base post payload).
rawData: true	Rows are the plain base post objects, byte-identical to the base Facebook Posts Scraper. No creative block, no filtering.

Input

Field	Type	Default	Description
startUrls	array	—	Facebook page URLs or usernames. Required.
resultsLimit	integer	20	Max posts per page.
captionText	boolean	true	Include the post caption text (OCR/entity extraction runs over it).
onlyPostsNewerThan	string	—	Absolute (2024-01-01) or relative (7 days) start date.
onlyPostsOlderThan	string	—	Absolute or relative end date.
includeMedia	boolean	true	Build the per-post image manifest.
extractEntities	boolean	true	Extract links / hashtags / prices / CTAs / etc.
ocrKeywords	array	[]	Keep only posts whose text or image-OCR contains any of these.
mediaManifest	boolean	true	Save a run-level media manifest to the key-value store.
metricsOnly	boolean	false	Drop base post fields, keep only creative.
rawData	boolean	false	Return the plain base post array unchanged.
aiEnhancement	boolean	false	Add AI sentiment / topics / brand-safety over the creative text.
aiModel	enum	claude-haiku-4-5	AI model; provider auto-detected from the name.
aiApiKey	secret	—	API key for the AI provider (falls back to the provider env var).
proxyConfiguration	object	Apify proxy	Residential → datacenter fallback, or custom proxy URLs.

How it works

The scraping core discovers the page id and GraphQL doc_id from the live page, then fetches the timeline feed with cursor-based pagination — the identical, unmodified core the base scraper uses (so rawData output is byte-for-byte the base output). The Media & OCR feature layer runs entirely on the already-fetched data: it reads media[].photo_image, media[].accessibility_caption, media[].id, and the post text, and reshapes them into the image manifest, OCR text, and entity lists. No additional Facebook network calls are made by the feature layer.

Keywords

facebook posts scraper, facebook image extractor, facebook OCR, image text extraction, accessibility caption, facebook media manifest, facebook creative harvester, hashtag extraction, link extraction, CTA detection, facebook page monitoring, ad creative, brand safety, social media image download, facebook post entities.

Legal & compliance

This actor collects only publicly accessible data and does not bypass login restrictions or access private content. Use scraped data responsibly and in compliance with applicable laws and platform terms (GDPR, CCPA).