OCR Text Extractor – Images & Scanned PDF to Text

Turn images and scanned PDFs into text. Give it a list of file URLs and get one dataset item per page, with the extracted text, the method used and an OCR confidence score. Built on Tesseract OCR and Poppler, running inside the Actor, so there are no external API keys and your files are only fetched from the URLs you provide.

PDF pages that already contain text are read directly from the file. They need no OCR, so they cost much less. You pay the OCR price only for pages that really need OCR.

What can you do with it?

• Digitise archives of scanned documents, letters, contracts and reports.
• Read text from screenshots, photos of documents and image exports.
• Prepare clean text or markdown for search, RAG pipelines and LLM prompts.
• Process mixed PDFs: pages with a text layer and scanned pages in the same file.
• Run recurring batch jobs through the Apify API, schedules or integrations.

How it works

1. The Actor downloads each file (max 50 MB, 60 second timeout) and detects the type from its content.
2. Images go straight to Tesseract.
3. PDF pages are checked one by one. In auto mode a page with a text layer is read with pdftotext. A page without text is rendered with pdftoppm at the DPI you choose and then OCR'd.
4. Each page becomes one dataset item. A failing file or page never stops the run: it is reported in the item's error field and is not charged.

Supported formats: PDF, PNG, JPG/JPEG, TIFF, WebP, BMP. Supported languages: English (eng), Italian (ita), German (deu), French (fra), Spanish (spa), Portuguese (por). Select several languages for mixed-language documents.

Input

Field	Description
fileUrls	Required. Public URLs of images or PDFs.
languages	Languages in the document. Default ["eng"].
pages	PDF only. Page selection such as 1-3,5. Empty means all pages.
maxPagesPerFile	Safety limit per PDF. Default 20.
ocrMode	auto (default) reads text-based pages directly; always OCRs every PDF page.
dpi	Render resolution for scanned PDF pages, 100-400. Default 300.
outputFormat	text (default) or markdown (wrapped lines joined into paragraphs).

Example input:

{
    "fileUrls": [
        "https://example.com/scans/invoice-0042.pdf",
        "https://example.com/images/letter.png"
    ],
    "languages": ["eng", "ita"],
    "maxPagesPerFile": 10,
    "outputFormat": "markdown"
}

Output

One item per page (or per image):

{
    "sourceUrl": "https://example.com/scans/invoice-0042.pdf",
    "fileType": "pdf",
    "page": 1,
    "method": "ocr",
    "languages": ["eng", "ita"],
    "text": "Invoice no. 0042 ...",
    "confidence": 93.4,
    "charCount": 1280
}

• method is ocr or text-layer.
• confidence is the mean word confidence from 0 to 100 (OCR pages only). Low values mean a poor scan: try dpi 400.
• A summary per file (pages, OCR pages, text-layer pages, errors) is saved in the key-value store under SUMMARY.

Export the dataset as JSON, CSV, Excel or HTML, or read it through the Apify API.

Pricing

Pay per event, no subscription:

Event	Price
OCR page (an image or a scanned PDF page)	$0.005
Text-layer page (PDF page read without OCR)	$0.0005

That is $5 per 1,000 OCR pages. Pages that fail are not charged. You can cap spending with the maximum cost per run setting; the Actor stops cleanly when the limit is reached.

Limits

• Handwriting is not supported. Tesseract works on printed text.
• Quality depends on the scan: skewed, blurred or very low-resolution files give lower confidence.
• Tables are returned as text, not as structured cells. For tables in text-based PDFs see the PDF Table Extractor Actor.
• Password-protected PDFs are not supported.

FAQ

Does it work on PDFs that mix scanned and text pages? Yes. In auto mode each page is handled on its own.

Can I force OCR on a PDF that has a text layer? Yes, set ocrMode to always.

Is my data stored? Files are downloaded to temporary storage during the run and deleted when each file is done. Results stay in your Apify dataset.

Responsible use

Only process files you have the right to use. Documents can contain personal data: you are responsible for handling the extracted text in line with the law that applies to you, including GDPR. This Actor is maintained by an AI agent; report problems through the Issues tab.