OCR — Scanned PDF & Image to Text (Tesseract, 29 Languages)

Extract text from scanned PDFs and images (PNG, JPEG, TIFF, WebP, GIF, BMP) with Tesseract 5 OCR in 29 languages. PDF pages that already have text are read directly at a tenth of the price. Per-page text and confidence, optional word bounding boxes. $3 per 1,000 OCR pages.

Extract text from scanned PDFs and images with OCR. Paste links to scanned documents, photos of pages, screenshots, faxes or receipts (PDF, PNG, JPEG, TIFF, WebP, GIF, BMP, JPEG 2000) and get the text back, page by page, with an OCR confidence score for every OCR'd page. Powered by Tesseract 5 with its LSTM neural-network models in 29 languages (English, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Nordic languages, Czech, Slovak, Hungarian, Romanian, Turkish, Russian, Ukrainian, Arabic, Hebrew, Hindi, Chinese, Japanese, Korean, Vietnamese, Indonesian, Thai), up to 4 at once. Smart PDF mode reads pages that already contain text straight from the text layer (exact, and a tenth of the price) and only OCRs the scanned pages, so mixed PDFs cost what they should. Optional word-level bounding boxes with per-word confidence for redaction, highlighting and form extraction. Use it to make scanned archives searchable, feed RAG pipelines and LLMs with paper documents, digitise invoices, contracts and letters, or give AI agents a reliable OCR tool. Simple pricing: $3 per 1,000 OCR pages (scanned-PDF OCR elsewhere costs $15 per 1,000) and $0.30 per 1,000 text-layer pages, with no compute bill on top. Failed downloads, unreadable files and blank pages are never charged.

What data does it return?

Every run outputs clean, schema-validated rows like these:

url	fileType	pagesOcr	confidence	wordCount	text	pagesTextLayer
https://raw.githubusercontent.com/ocrmypdf/OCRmyPDF/main/tests/resources/linn.pdf	pdf	1	95.5	728	The LinnSequencer\n32 Track MIDI Sequence Recorder\n\nThe LinnSequencer is a state-of-the-art composition and performance tool…
https://raw.githubusercontent.com/tesseract-ocr/test/main/testing/phototest.tif	tiff	1	96.3	60	This is a lot of 12 point text to test the\nocr code and see if it works on all types\nof file format.
https://arxiv.org/pdf/1706.03762	pdf	0		6457	Provided proper attribution is provided, Google hereby grants permission to reproduce the tables and figures…	15

How much does it cost?

Transparent pay-per-event pricing — you only pay for what you get, and the actor always respects your run budget: set a maximum total charge and it stops cleanly at the cap instead of overrunning.

Event	Price	Charged	Worked example
page-ocr	$0.003	per image or scanned PDF page read by OCR	A 10-page scanned PDF = $0.03; 1,000 receipts = $3.00
page-text	$0.0003	per PDF page that already has a text layer (read directly, no OCR)	A 100-page digital PDF = $0.03
apify-actor-start	$0.00005	per run per GB of memory (platform start fee)	One run at 4 GB = $0.0002

Input options

Field	Type	Required	Default	Description
urls	array	yes	—	Links to scanned PDFs or images (PNG, JPEG, TIFF, WebP, GIF, BMP, JPEG 2000). Each image is one page.
languages	array	no	["eng"]	Languages in the documents (up to 4). Picking only the languages you need is faster and more accurate.
pdfMode	string	no	"auto"	Auto: pages that already contain text are read directly (exact, and charged at the cheaper text-layer price); only scanned pages are OCR'd. OCR all: OCR every page, e.g. when a PDF's hidden text layer is poor.
outputMode	string	no	"document"	For PDFs: one row per PDF with all pages, or one row per page. Images are always one row each.
includeWords	boolean	no	false	Add every recognised word with its confidence and pixel box (left, top, width, height), e.g. for redaction, highlighting or form extraction.
minWordConfidence	integer	no	0	Drop words Tesseract is less sure of than this (0–100). 0 keeps everything; 40–60 removes most noise from stains and stamps.
pageSegmentation	string	no	"auto"	How Tesseract finds text. Auto suits most pages; Single column for uniform text; Single block for receipts and snippets; Sparse for scattered text like forms and screenshots.
dpi	integer	no	300	Resolution scanned PDF pages are rendered at before OCR. 300 is the accuracy sweet spot; 200 is faster for large print.
maxPagesPerPdf	integer	no	200	Process at most this many pages of each PDF (from the start). Longer PDFs are marked truncated.
maxFiles	integer	no	1000	Stop after this many files.
password	string	no	—	For password-protected PDFs you are entitled to open. Used for every PDF in the run.

Limits & honest notes

• Tesseract reads printed and typed text very well (95%+ confidence on clean scans); handwriting, low-resolution photos, heavy stamps and decorative fonts come out poorly. The confidence score tells you which pages to check.
• Output is plain text in reading order with line and paragraph breaks. Tables come out as lines of text, not structured tables. Images and charts have no text to extract.
• Pages rotated by 90° or 180° are not auto-rotated; slightly skewed scans are fine.
• Up to 4 languages per run; picking only the languages your documents use is faster and more accurate. Modern Greek is not available.
• HEIC/AVIF photos and SVG are not supported (convert photos to JPEG). Files up to 100 MB; PDFs up to 2,000 pages each (200 by default).
• Password-protected PDFs need the password input. Links that return a web page instead of a file are reported and not charged.
• Respects your run budget: set a maximum total charge and processing stops cleanly at the cap, with no page charged twice.

Related ActorForge actors

• PDF OCR
• Image to Text OCR
• PDF to Markdown — for digital PDFs with a text layer
• Word, Excel & PowerPoint to Markdown

Changelog

• 2026-10-19 — Initial release: Tesseract 5 OCR (fast LSTM models) for images and scanned PDFs in 29 languages, smart text-layer mode for PDFs, per-page confidence, optional word boxes.

---

Built and maintained by ActorForge. Reliability first: every actor in the portfolio is monitored by an automated nightly self-test, and breakages are typically fixed the same day.