Image OCR: Extract Text from Images, Screenshots & PDFs avatar

Image OCR: Extract Text from Images, Screenshots & PDFs

Pricing

from $1.78 / 1,000 image or page ocreds

Go to Apify Store
Image OCR: Extract Text from Images, Screenshots & PDFs

Image OCR: Extract Text from Images, Screenshots & PDFs

Image OCR in bulk: extract text from image and PDF URLs, with per-line boxes and confidence. Screenshots, receipts, scans, scanned PDFs. No API key, no proxy.

Pricing

from $1.78 / 1,000 image or page ocreds

Rating

0.0

(0)

Developer

Detroit Web Agent - Open Data

Detroit Web Agent - Open Data

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

8 days ago

Last modified

Features

Limited permissions

Share

Image OCR in bulk. Give it links to images or PDFs and get the text back, plus every line with its box and a confidence score. It reads printed text in screenshots, receipts, invoices, scanned pages, signs and photos of documents.

The OCR runs inside the Actor on CPU. No API key, no external service, no proxy and no per-request vision model bill.

Run in 30 seconds

  1. Open the Input tab. The prefilled test lists 26 small receipts, invoices and notes.
  2. Click Start.
  3. Open the Text table in the Output tab, or download the dataset as JSON, CSV or Excel.
  4. Replace the list with your own image and PDF links and run again.

What does this Actor do?

For each image or PDF link it downloads the file, finds the text and returns one row per image or PDF page:

GroupFields
Sourceurl, fileType, page, pageCount
Texttext (the whole page, lines in reading order), lineCount, wordCount
Detaillines: each line with text, confidence and four-corner box
QualitymeanConfidence (0 to 1)
Imagewidth, height, method (ocr or text-layer), durationMs
Statuserror (why a file was skipped), extractedAt

PDF pages that already contain real text are read directly from the text layer, which is exact and cheaper. Scanned PDF pages are rendered at 200 dpi and read with OCR.

Limits

  • It reads printed text well. Handwriting, very small or blurry text and heavy decoration are unreliable. Check meanConfidence and the lines boxes to spot weak pages.
  • Languages: English and other Latin-alphabet text, plus Chinese. Right-to-left scripts, Japanese kana and Korean are not supported in this version.
  • Files must be reachable by a public link and no larger than 25 MB. Very large images are reduced to 3,200 pixels on the long side before reading; boxes are mapped back to the original size.
  • Animated images: only the first frame is read.
  • PDFs: the first 20 pages by default, up to 200 with Max pages per PDF.
  • Tables are returned as lines in reading order, not as rebuilt cells.
  • If a run reaches its time limit it stops cleanly, keeps what was read and lists the unread links in OUTPUT.

Input

FieldWhat it does
fileUrlsImage or PDF links, one per line.
outputFormatboth (text and lines), text or lines.
maxPagesPerPdfPages read from each PDF. Default 20.
maxFilesStop after this many files.

Example input

{
"fileUrls": [
"https://example.com/receipt.jpg",
"https://example.com/contract-scan.pdf"
],
"outputFormat": "both",
"maxPagesPerPdf": 10
}

Output sample

{
"url": "https://example.com/receipt.jpg",
"fileType": "jpeg",
"page": 1,
"pageCount": 1,
"text": "Maple Street Market\nReceipt #4000\nPaper towels $11.07\nTOTAL $28.16",
"lines": [
{"text": "Maple Street Market", "confidence": 0.997, "box": [[48, 40], [420, 40], [420, 82], [48, 82]]}
],
"lineCount": 4,
"wordCount": 11,
"meanConfidence": 0.996,
"width": 640,
"height": 292,
"method": "ocr",
"durationMs": 1200,
"error": null,
"extractedAt": "2026-11-10T03:00:00.000Z"
}

Your own run returns the real values for every field.

Use cases

  • Pull amounts, dates and vendor names out of receipts and invoices for bookkeeping.
  • Turn screenshots of chats, dashboards or tables into searchable text.
  • Make scanned PDF archives searchable and feed them to a language model.
  • Check large sets of product or ad images for text.

Tips for good results

  • Send the sharpest version of the image you have. Text at least 20 pixels tall reads best.
  • Use lines to find where on the page a value sits using the box corners.
  • Filter on meanConfidence below 0.8 to queue weak pages for a human check.
  • Use outputFormat: text for the smallest dataset.

Field reference

meanConfidence is the average of the line confidences from the OCR model. method is ocr for images and scanned pages and text-layer for PDF pages with embedded text. box lists four corners (top-left, top-right, bottom-right, bottom-left) in the original image pixels. error is set when a file could not be downloaded or decoded; those rows are not charged.

Pricing and cost control

Pay per page: $0.0029 per image or OCRed PDF page on the free plan, $0.00178 on the Gold plan (Bronze $0.0027, Silver $0.00231), and $0.000483 per PDF page read from its text layer ($0.000297 on Gold), plus a $0.00005 start fee per run. That is $2.90 per 1,000 images. Files that fail are not charged. Use Maximum cost per run or Max files to cap spend; the run stops cleanly at the limit.

Integrations and scheduling

Send results to Google Sheets, Slack, email, Make, Zapier or n8n with the Apify integrations, or call the Actor from code through the REST API and the Python and JavaScript clients. Open the API tab for ready-to-copy snippets.

API (one call, returns the rows as JSON):

$curl -X POST "https://api.apify.com/v2/acts/transparent_meteorite~image-ocr-text-extractor/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" -H "Content-Type: application/json" -d '{"fileUrls":["https://api.apify.com/v2/key-value-stores/5htCX46pZv4alJ3U3/records/sample-01.png"]}'

Python

from apify_client import ApifyClient
client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("transparent_meteorite/image-ocr-text-extractor").call(run_input={"fileUrls":["https://api.apify.com/v2/key-value-stores/5htCX46pZv4alJ3U3/records/sample-01.png"]})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
print(row)

AI agents (MCP): connect the Apify MCP server at https://mcp.apify.com?tools=transparent_meteorite/image-ocr-text-extractor to Claude, Cursor or any MCP client, and your agent can run this Actor on request.

The Actor reads only files you point it to, at links you are allowed to access. It does not search the web or sign in anywhere. Make sure you have the right to process the documents you submit, especially documents with personal data. This is an independent tool.

FAQ

Do I need an API key? No.

Can I upload files directly? Use a link. A file stored in an Apify key-value store has a link you can paste.

Why is the text slightly wrong? Low resolution, unusual fonts and photos taken at an angle lower accuracy. Check meanConfidence.

Does it read tables? It reads the text in them, line by line, with positions. It does not rebuild cells.

Are failed files charged? No.

Support

Found a bug or a file that reads badly? Open an issue on the Issues tab with the run URL and the file link.

Changelog

  • 0.1: First private build. Images and PDFs, per-line boxes and confidence, text-layer shortcut for PDFs.