Cursive Translator

Handwriting transcription desk

Cursive translator: upload a picture of cursive handwriting

Submit a handwriting image for machine transcription. When the service returns text, it is written into the editable box on the right for you to review against the image. The request can also come back empty or as an error, and the page prints whichever it receives in the status line.

There is no separate “cursive decoder” or “cursive decipher” mode on this site. Both labels lead to this same upload-and-review flow: one image goes to the same endpoint, and the returned text appears in the same editable box.

Site policy on what comes back: it is machine output, not a reading of your document. Nothing in the box marks which words matched your page. Compare it against the original before you copy, publish or act on it.

1. Add an image of the handwriting

JPG, PNG, WEBP, GIF, BMP or TIFF, up to 6 MB. You can also paste an image from your clipboard while focus is outside an editable control.

The box below is editable. Copy and Download act on what is in the box when you press them, not on what was returned.

Where the image goes. Pressing Read cursive reads the file in your browser, posts it to /api/cursive-to-text on this site, and that endpoint sends it to Google Cloud Vision and returns the text field of the response. The site code does not write your image or the returned text to any store of its own. Site policy: review whatever comes back against the image before you quote, publish or act on it, and do not upload anything you are unwilling to send to an external processing service.

What the page does when you press the button

Three steps, no account, nothing to install.

  1. Add one image

    Choose a file, drag one onto the box, or press Ctrl or Cmd + V to paste one. The page checks the type and the size first: anything outside JPG, PNG, WEBP, GIF, BMP or TIFF, or larger than 6 MB, is refused in your browser and never sent. Submit one page per request; the form takes a single image.

  2. Press Read cursive

    The file is read as base64 in your browser and posted to /api/cursive-to-text. That endpoint sends one Google Cloud Vision DOCUMENT_TEXT_DETECTION request with the language hint en, and returns the text field of the response. The status line reports what came back — text, an empty result, or an error.

  3. Review, then keep it

    Edit the box against the image, then press Copy text or Download .txt. The script assigns the download anchor your image filename with a .txt extension, or handwriting.txt if the selected file has no usable name; whether and where it is saved depends on the browser.

Capture checklist

Instructions for taking the image, and one experiment you can run yourself. None of these is a claim about what the service will return.

  • Light the page evenly and keep your own shadow off it. Step to the side rather than leaning over the page.
  • Hold the camera parallel to the paper, so the written lines run horizontally across the frame.
  • Flatten the paper: press out folds, and hold curved book pages down at the edges.
  • Fill the frame with the writing. Leave out the desk, your fingers and the facing page.
  • Zoom to full size and read the image yourself before uploading. Anything you cannot read on screen, you will not be able to check against the returned text.
  • Submit one page, column or paragraph per request rather than a whole spread.
  • Upload the original file rather than a screenshot of it, and keep the original: it is your reference for the review pass.
  • An experiment, not a rule: submit the same page twice — once as shot and once re-shot with a change you control, such as different light, a straighter angle or a tighter crop — and compare the two returned texts against the page yourself. That comparison tells you about your page; this site makes no prediction about how it will come out.

The review pass

This client reads only the returned text field and writes that block into the box; it does not display confidence values, highlighting or other response fields. Nothing on this page tells you which words match your document. This is a checklist for comparing the box with the image.

  • Read the box against the image, line by line. Put them side by side and check word against word rather than skimming for obvious errors.
  • Check each name, place, date and figure against the page character by character, and against another document where you have one.
  • Where a letter is in doubt, find the same letter elsewhere in the same document and compare the two shapes directly. Write down which one you decided on and why.
  • Ask what else the word could be. Substitute each candidate reading into the sentence and see which ones produce a word you can confirm from the image.
  • Mark, do not guess. Type [?] in the box wherever you cannot resolve a word from the image, instead of leaving a plausible word standing unchecked.
  • Try a re-submission as an experiment. Crop one difficult paragraph on its own, submit that, and compare the two returned texts against the page yourself. Nothing here predicts what either comparison will show.
  • Keep the image with the text if the page matters, so you retain the source used for review.

Configured checks, and questions for your review

The left column describes checks in the current client and endpoint code. The request thresholds are best-effort controls, not hard platform quotas. The right column is a worksheet for your page.

Checks configured in the code

  • Client-accepted types: JPG, PNG, WEBP, GIF, BMP, TIFF. Anything else is refused in your browser with a message in the status line. The endpoint does not repeat this MIME-type check; it checks for a nonempty encoded image string and its encoded length.
  • Size cap: 6 MB. Larger files are refused before any request is made, and the status line states the file's size.
  • One image per request. There is no queue, no batch and no multi-page upload.
  • In-memory rejection branch: above 10 requests per IP in 60 seconds within one module isolate. This is not a stable distributed rate limit.
  • KV rejection thresholds: 100 requests per IP and 3000 site-wide per UTC day. The counters are fail-open and use non-atomic read-modify-write operations, so these are best-effort checks rather than hard caps.
  • One language hint is sent with every request: en. It is fixed and there is no control for it.

Questions to answer from your own page

  • Can you read every word of the image yourself at full zoom? List the ones you cannot.
  • Is the ink continuous, or broken by fading, bleed-through, staining or crossing-out? Mark where.
  • Does any part of the page carry meaning by position — columns, tick boxes, a table, a note in the margin? Write down what would be lost if it arrived as running text.
  • Which words are names, places, dates or figures? Those are the ones you will have to check one by one.
  • Is any of it in a language other than English, or in an unfamiliar letterform? Note which lines.
  • What would a wrong word here cost you? Set your review depth from that answer.

Answer the right-hand column before you submit and again after the text comes back: the answers are your review list. The worksheets linked below set out procedures for working through a page by eye.

Questions

Is this cursive translator free?

There is no account, sign-up or payment control on this site. The endpoint has KV rejection thresholds of 100 requests per IP address and 3000 across the site per UTC day. Those counters fail open and update non-atomically, so they are best-effort controls rather than hard caps. The page prints any returned rejection message in its status line.

How accurate is it?

This site publishes no accuracy figure and makes no claim about how any document will come out. This client reads only the returned text field and displays no confidence values, so neither the page nor this answer can tell you which returned words match your page. Compare the box against the image yourself.

What happens to my image?

Your browser posts it to /api/cursive-to-text on this site; that endpoint sends it to Google Cloud Vision and returns the text field of the response. The site code has no image or transcript database, does not intentionally persist either item, and provides no saved-work or history retrieval feature. This page does not describe what Google logs, retains or transmits; consult Google's own current policies for that information.

Which file types and sizes are accepted?

JPG, PNG, WEBP, GIF, BMP and TIFF, up to 6 MB per image. Both checks run in your browser before anything is sent, and the status line states which one refused the file.

Can it translate cursive into another language?

No. There is no translation step anywhere in the code: the request asks for text recognition with the language hint en and returns text. Paste the result into a translation service yourself if you need another language.

Is there a setting for old documents, or for another language?

No. All three upload pages send the image with the same feature type and the same single language hint. There is no model choice, script setting or per-page OCR tuning; ordinary request metadata and counter state can still differ.

Can I use it on a phone?

The page has no desktop-only code, and the file input is a standard one, so a mobile browser offers whatever sources it normally offers for choosing an image. What those are is your browser's behaviour, not this site's.

Does it work on typed or printed text?

The request is sent whatever image you choose; nothing in the code distinguishes handwriting from print, and the same endpoint handles both. Submit your page and compare what comes back against it. The handwriting to text page runs the same request with copy written for notes and forms.

Everything on this site