Image to Text — OCR for Screenshots, Scans and Photos

Convert image to text with OCR: screenshots, scanned documents, receipts, PNG, JPG, WEBP, TIFF, HEIC and scanned PDFs, in 19 languages. Every image comes back as plain text with an engine confidence score, and every word can come back with its position on the image. $0.004 per image, no fee per run, no API key, no external OCR service.

It is the OCR Text Extractor engine with the form set up for images; anything that works there works here.

What it is for

What you give it	What you get back
A screenshot of a web page, app or chat	The text on it, line by line
A scanned letter, form or invoice	Recognised text with a confidence score
A batch of hundreds of image links	One row per image, same fields for all of them
A scanned PDF	Text page by page, pages that already contain text copied for free
Base64 data from your code or an AI agent	The same row, nothing to host

An empty call ({}) reads a sample invoice image, so you can see the output shape before you send your own files.

What data does it return?

Field	Type	Example
source	string	https://example.com/receipt.jpg
fileName	string	receipt.jpg
format	string	image or pdf
language	string	eng
text	string	INVOICE 2026-114\nMalina Data…
wordCount / charCount	integer	26 / 177
confidence	number	93.7 (average, 0–100)
pages	array	[{"number": 1, "text": "…", "confidence": 93.7, "source": "ocr"}]
words	array	[{"text": "INVOICE", "confidence": 95, "left": 124, "top": 168, "width": 174, "height": 32}] (when switched on)
status / errorCode / errorMessage	string	ok, or why a file failed
extractedAt	string	2026-10-06T02:25:40+00:00

How much does it cost?

Event	Price	When it is charged
Image or page read with OCR	$0.003	Per image, or per scanned PDF page, that returned text
File processed	$0.001	Once per file that returned text

So one image is $0.004 and 1,000 images are $4. There is no fee per run. Files that fail, or in which no text was found, come back as error rows and are never charged. PDF pages that already have a text layer are copied, not OCR-ed, and are not charged as OCR pages.

Job	Cost
1 screenshot	$0.004
1,000 product photos or receipts	$4.00
100 scanned PDFs, 10 pages each	$3.10

The Apify Free plan includes $5 of monthly usage — about 1,250 images.

How to turn an image into text

1. Click Try for free and sign in to Apify.
2. Paste image links into Image URLs, or use Upload a file.
3. Pick the OCR language — the right language matters more than any other setting.
4. Press Start, then read the Output tab or export JSON, CSV or Excel.

⬇️ Input

{
  "urls": [
    "https://raw.githubusercontent.com/lergassy/apify-actor-assets/main/document-text-extractor/sample-scan.png"
  ],
  "language": "eng",
  "includeWordBoxes": false
}

Field	What it does
urls	Direct links to images or PDFs. The link must open the file itself, not a gallery page
file	Upload one file from your computer
base64Files	Base64 strings or {"fileName", "data"} objects, for API and agent calls
language	One of 19 languages: English, German, French, Spanish, Portuguese, Italian, Dutch, Polish, Russian, Ukrainian, Turkish, Indonesian, Vietnamese, Arabic, Hindi, Chinese (simplified and traditional), Japanese, Korean
autoDetectLanguage	Inspect the image first and switch language when the script is clearly Cyrillic, Arabic, CJK or Devanagari. Works on documents; on street photos it can guess wrong, so set the language yourself when you know it
includePages	Per-page text and confidence (on by default)
includeWordBoxes	Every word with its confidence and pixel box
minWordConfidence	Drop words the engine is less sure about than this (0–100). 60 removes most noise from phone photos
useExistingTextLayer	For PDFs: copy text that is already there instead of OCR-ing it (on by default)
pageRange, maxPagesPerFile	Which PDF pages to read, and a safety cap
maxConcurrency, maxFileSizeMb, timeoutPerFileSecs	Advanced limits

If a host refuses the download (rate limit, block), the Actor retries with a pause and then through a residential proxy before giving up on the file. You are not charged for that.

⬆️ Output

{
  "source": "https://…/sample-scan.png",
  "format": "image",
  "language": "eng",
  "text": "INVOICE 2026-114\nMalina Data\nJalan Raya Uluwatu 88, Jimbaran, Bali\nConsulting services 1,250.00 USD\nDocument processing 340.00 USD\nTotal due: 1,590.00 USD\nPayment terms: 30 days",
  "wordCount": 26,
  "confidence": 93.7,
  "pages": [{ "number": 1, "text": "INVOICE 2026-114…", "confidence": 93.7, "wordCount": 26, "source": "ocr" }],
  "status": "ok"
}

Error rows

errorCode	Meaning
http-error / network-error / timeout	The file could not be downloaded, even after retries
too-large	Above the size limit (50 MB by default, up to 200 MB)
unsupported-format	Not an image or a PDF — for example a link to a web page
corrupt-file	The file is damaged
no-text-found	Nothing readable — usually the wrong language, a photo with very little text, or a very small image

Limitations, stated plainly

• Best on printed text: screenshots, scans, typed letters, invoices. Screenshots and clean scans typically read at 85–95% confidence.
• Phone photos of receipts, signs and labels read noticeably worse (40–70% confidence in our tests), and a photo where the text is a small part of the scene may return nothing. Google Lens–based Actors handle such photos better; this one is cheaper and tells you how sure it is.
• Handwriting is not read reliably.
• One very dense page (an old newspaper page with ~2,000 words) can take over a minute.
• Recognition runs inside the Actor (Tesseract), so there is no external quota — and no cloud AI either.

Use cases

• Screenshots to text — pull text out of app screens, chats, dashboards or error messages.
• Digitising paperwork — invoices, permits, letters and records that only exist as scans, with a confidence score that tells you which need a human.
• Feeding images to an LLM or a search index — pass text instead of pictures, with noisy words filtered out.
• Document parsers — word boxes let you read a field by where it sits on the page.

🤖 For AI agents and LLM apps

{ "urls": ["https://example.com/screenshot.png"], "language": "eng" }

• One row per file; status is ok or error, no log parsing needed.
• confidence gives the agent a reason to ask for a better image instead of acting on noise.
• Files can be passed as base64.
• An empty call returns a sample result instead of an error.

Use via API

curl -X POST "https://api.apify.com/v2/acts/lergassy~image-to-text/run-sync-get-dataset-items?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"urls": ["https://example.com/scan.png"], "language": "eng"}'

from apify_client import ApifyClient
items = ApifyClient("YOUR_TOKEN").actor("lergassy/image-to-text").call(
    run_input={"urls": ["https://example.com/scan.png"]}
).iterate_items()

Works the same from JavaScript, n8n, Make, Zapier and the Apify MCP server.

❓ FAQ

Do I need an API key?

No. The OCR engine runs inside the Actor.

Which image formats work?

PNG, JPG, WEBP, TIFF, BMP, GIF and HEIC (iPhone photos), plus PDF.

What is the difference from OCR Text Extractor?

None in the engine. OCR Text Extractor is the main Actor, framed around scanned PDFs; this one is set up for images.

Can it read a PDF that already has text?

Yes, but Document Text Extractor is the better tool for that: it returns Markdown, tables and RAG chunks.

You might also like

Actor	What it does
OCR Text Extractor	The main Actor: the same engine, framed around scanned PDFs
PDF Text Extractor	Text, Markdown and tables from PDF files
Document Text Extractor	PDF, Word, Excel and PowerPoint to Markdown and RAG chunks
Speech to Text	Audio and video to a timed transcript

Also known as

Image to text converter, OCR API, picture to text, photo to text, screenshot to text, extract text from image, JPG to text, PNG to text.