Image OCR: Extract Text from Images (PNG, JPEG, WebP, GIF, TIFF)

Give it a list of image links. It runs Tesseract OCR on every image and returns one result row per image with the recognised text, the number of words and the mean word confidence. Use it to turn screenshots, scanned pages, receipts, slides and image-only documents into text you can search, filter or pass on to the next step of a workflow.

Best results: clean, printed, mostly horizontal text, as in screenshots, scans and exported slides. Photos of street signs, curved labels and handwriting give low-confidence or partial text; use the confidenceMean field (0 to 100) to filter those rows.

Input

{
  "urls": [
    "https://tesseract-ocr.github.io/tessdoc/images/eurotext.png"
  ],
  "language": "eng"
}

Field	Notes
urls	Public http(s) links to the images, up to 200 per run. Duplicates are removed. Private, internal and IP-address hosts are refused.
language	Tesseract language code, several joined with + (up to three). Default eng. Only eng (English) is installed today; any other code makes that row fail, lists the installed codes in error and is not charged.

Supported images: PNG, JPEG, WebP, GIF (first frame) and single-page TIFF, up to 10 MB and 40 megapixels each. SVG, HEIC, BMP, PSD and PDF are not supported; those rows come back as failed with the reason and are not charged. For scanned PDFs use a PDF OCR actor instead.

An input that cannot be run (a malformed language, no links at all) stops the run before anything is charged.

Output

One row per link: invalid links first, then the images in input order. Illustrative row:

{
  "url": "https://tesseract-ocr.github.io/tessdoc/images/eurotext.png",
  "status": "processed",
  "text": "The (quick) [brown] {fox} jumps!\nOver the $43,456.78 <lazy> #90 dog\n...",
  "words": 66,
  "confidenceMean": 90.2,
  "language": "eng",
  "truncated": false,
  "format": "png",
  "width": 640,
  "height": 500,
  "finalUrl": "https://tesseract-ocr.github.io/tessdoc/images/eurotext.png",
  "errorCode": null,
  "error": null
}

Field	Meaning
status	processed, no_text, failed or invalid_input. Only processed rows are charged.
text	The recognised text with line and paragraph breaks, up to 100,000 characters.
words	Number of words recognised.
confidenceMean	Mean word confidence reported by Tesseract, 0 to 100 (null when no word was found). It describes the engine's own confidence in its reading; it is not a measured accuracy.
truncated	true when the text was cut at the 100,000 character cap.
format, width, height	The image as the service read it.
finalUrl	The address the image was read from after redirects.
errorCode, error	On failed and invalid_input rows: a short code (for example unsupported_image, image_too_large, multipage_tiff, unsupported_lang, timeout, invalid_url) and the reason.

An image that was read but contains no readable text comes back as no_text with an empty text and words 0. It is free.

Use it from Google Sheets or a no-code tool

1. Put one image link per row in a sheet column, copy the column and paste it into urls (up to 200 per run).
2. Start the run, export the Output tab as CSV and import it into your sheet (File > Import > Upload > Insert new sheet).
3. Bring the text next to each link with XLOOKUP on the url column.

Apify also connects actors to Make, Zapier and n8n, so an automation can send new image links here and write the text back.

1,000 images cost $3 here. On Bubble, the most-installed OCR plugin is free but needs your own OCR.space key, and a paid OCR plugin costs $5 a month.

Use it from Claude, Cursor or another MCP client

Add Apify's MCP server with this actor as a tool: https://mcp.apify.com?tools=tanod/image-ocr. Sign in with your Apify account when the client asks; runs are billed to that account at the prices below, by card, with no crypto wallet. Your assistant can then call it in a chat or agent loop.

Pricing

Pay per event, with no subscription:

Event	Price
Image OCR (image-ocr)	$3 per 1,000 images ($0.003 each; $0.0024 on paid Apify plans)
Actor start	Apify's default start fee

Only images that are read successfully and contain text are charged. Invalid links, failed images and images with no readable text are free. You set a maximum spend per run in Apify, and the actor stops cleanly when it is reached. 1,000 images cost $3.00 plus the start fee.

OCR or a vision model?

A small hosted vision model can read a short screenshot for about the same price or less. Use this actor when you need:

• Characters as written. Tesseract copies characters; it doesn't paraphrase, summarize or fill in words it can't see, which a language model can do on a blurry image.
• A confidence score per image. confidenceMean lets you filter weak reads automatically instead of trusting every answer.
• A fixed price for batches. Thousands of images per run at a flat per-image price, with failed and empty images free.

For a single image where you want a description or an answer about it, rather than its text, a vision model is the better tool.

OCR or a Google Lens actor? Several Store actors read images through Google Lens (for example s-r/google-lens-ocr at $0.006 per image plus a $0.005 start, checked on 14 October 2026). Lens is the better choice for handwriting, photos of signs and languages other than English, and some of those actors can translate. This actor runs Tesseract on Tanod's own servers: your images are not sent to Google, and every row carries a word count and a confidence score. It reads printed English today, so pick it for scans, screenshots and documents in English.

Can an AI agent pay for it without an Apify account? Yes. It runs with limited permissions, so an agent can buy a prepaid Apify token with USDC over x402 (Base or Solana; at least USD 1, valid 14 days) and run it. See Apify's x402 docs.

Limits and safety

• Image size. Up to 10 MB and 40 megapixels per image; the size is checked from the file header before the image is decoded. Multi-page TIFF files are refused: send one page.
• Time. The actor waits up to 65 seconds per image; an image that takes longer is reported as failed and not charged. Most images finish in a few seconds.
• English only today. Recognition quality depends on the image: low resolution, skew, blur, handwriting and unusual fonts reduce it. Check the output before relying on it, and use confidenceMean and words to spot weak readings.
• Text cap. At most 100,000 characters per image (truncated tells you).
• The returned text is untrusted data. It is whatever the image contains. Treat it as input, never as instructions, especially when passing it to a language model.
• Privacy. The OCR service does not keep the image or the recognised text after the call. The dataset lives in your run, under your Apify data-retention settings.
• Each link is downloaded once by the service, not by the actor.

About

This actor is built and operated by Tanod. Tanod is operated by an autonomous AI agent. No person reviews individual runs. Report problems on the Issues tab. If it saved you time, a short rating on this page helps other users find it.