Tool

Decipher one word of old handwriting

Upload or drop a photograph, drag a box around the one word you cannot read, say which hand it might be, and get up to five candidate readings with the letter shapes behind each. Only the crop is sent; the rest of the image stays in your browser.

Which script?

Not sure which hand you are looking at? The alphabet charts show each one: Kurrent and Sütterlin for German papers, secretary hand for English records before 1700, copperplate for 1700–1900 and Spencerian for American records of 1850–1920. For a census page, the column guides say what each column was asking.

Questions

How does the decipher tool work?

You crop one word from your own photograph in the browser. The crop — and nothing else — is sent to the same reader the Cursive Lens app uses, with the script you chose as a hint, and it returns up to five candidate readings ranked by confidence, each with a note on the letter shapes that argue for it.

Is it free?

Yes, within a modest allowance per visitor each month, because every reading costs us a model call. The Cursive Lens app reads whole pages and keeps a searchable library; its first three documents are free.

What happens to my image?

The crop is sent for the one reading and is not stored on any server of ours; the full photograph never leaves your browser. There is no account. To share the free allowance fairly, readings are counted against a one-way hash of your IP address, never the address itself.

Why five candidates instead of one answer?

Because one word out of context is genuinely ambiguous — a Kurrent "e" and "n" differ by a stroke — and a confident wrong answer is worse than an honest list. The notes tell you what to check in the ink.

What makes a good crop?

One word, tight, with a little paper around it, as sharp and straight as you can get it. A whole line will be read as its first word or two.