Image to Text OCR - Extract Text From Images & Screenshots
Pricing
from $3.00 / 1,000 image reads
Image to Text OCR - Extract Text From Images & Screenshots
Extract the text from any list of image URLs: screenshots, scans, receipts, signs, slides. Returns the full text, a confidence score and word count per image, optionally every line with its position. 20 languages, long text split into agent-sized rows. Pay only for images with text.
Pricing
from $3.00 / 1,000 image reads
Rating
0.0
(0)
Developer
Kobkoon Janngam
Maintained by CommunityActor stats
0
Bookmarked
1
Total users
0
Monthly active users
2 hours ago
Last modified
Categories
Share
Image to Text OCR — extract text from images and screenshots
Read the text in any list of images. Paste image URLs (screenshots, scanned pages, receipts, slides, signs) and get one row per image: the full text, a confidence score and a word count. Optionally, every line comes with its position. Built for data pipelines and AI agents.
Quick start: press Start with no configuration. It reads eleven sample images in six formats, from a one-word logo to a newspaper column, so you can see every kind of row. Then paste your own links.
Sample output
| fileName | format | status | confidence | wordCount | text |
|---|---|---|---|---|---|
| testocr.png | png | ok | 94 | 60 | This is a lot of 12 point text to test the ocr code… |
| bill.png | png | ok | 83 | 97 | FIRST CHEQUING · Line of Credit 100,000.00 Rate 4.2… |
| cosmic.png | png | ok | 88 | 7 | HellO World from beyond the Cosmic Void |
| eurotext.tif | tiff | ok | 92 | 66 | The (quick) [brown] {fox} jumps! Over the $43,456.78… |
| simple.webp | webp | ok | 91 | 1 | Tesseract.js |
These rows come from the default run on 30 October 2026. They are real output, including the mistakes: "HellO" is how OCR read a decorative font.
Every row also carries url, lineCount, chunk/chunkCount, and a plain-English note when something went wrong. With Include each line with its position, each row also gets lines, a list of { text, confidence, box: { x, y, width, height } } entries.
What you can do with it
- Feed screenshots to an AI agent. It gets text instead of pixels, and each row fits in its context.
- Digitise scans and receipts. Turn a folder of image links into searchable text.
- Read text on product photos, menus and slides for search, translation or moderation.
- Find where text sits. Line boxes let you highlight it, redact it or crop around it.
Input
| Field | Default | Notes |
|---|---|---|
imageUrls | 11 samples | PNG, JPEG, WebP, GIF, BMP, TIFF or PNM. Up to 25 MB each and 500 per run. Each link must open without signing in. |
languages | eng | 20 languages, including Spanish, German, French, Japanese, Chinese, Arabic, Hindi and Thai. Combine them for mixed text. |
maxTextBytesPerRow | 3000 | Long text is split across rows at paragraph breaks, which suits agents and RAG pipelines. 0 keeps each image's text in one row. |
includeLines | false | Adds each line with its confidence and bounding box. |
Every image gets a row that says what happened
| status | Meaning | Charged |
|---|---|---|
ok | Text was found | Yes, once per image |
no-text | The image was read but contains no readable text | No |
not-an-image | The link is a PDF or a web page. Use pdf-text-extractor for PDFs | No |
download-failed | The link is dead or private, with the HTTP reason | No |
unreadable | The file could not be decoded | No |
One bad link never fails the run. The run fails only when none of the images could be read at all.
How it works
OCR is done by Tesseract, compiled to WebAssembly, and it runs inside the Actor. There is no browser, no API key and no third-party OCR service, and nothing leaves the run except the download of your own files. A slightly skewed scan is straightened before reading. A page that is turned on its side is not.
Accuracy depends on the image. Clean printed text at a normal size reads at 90%+ confidence. Handwriting, very small text and stylised fonts read poorly, and the confidence field tells you which rows to trust.
Pricing
You pay per image in which text was found. Images with no text and links that fail cost nothing.
Works well with
Same publisher, same output conventions, same honest-failure rules:
- pdf-text-extractor — Page-by-page text from any PDF, with OCR for scanned pages.
Disclaimer
This Actor reads only the files you link to: it scrapes no website and calls no third-party API — the OCR engine runs inside the Actor. Make sure you have the right to process the files you submit.
