Transcribe Handwritten Historical Documents with Transkribus

Go to transkribus.eu and click Sign Up. Enter your name, email, and create a password.

Check your email for the verification link and confirm your account.

1. Go to transkribus.eu
2. Click Sign Up in the top right corner
3. Fill in your name, institutional email, and password
4. Click the verification link in the confirmation email

What you should see: You're taken to the Transkribus dashboard, which shows "My Collections" with an empty list.

Troubleshooting: If the email doesn't arrive in 5 minutes, check your spam folder. Academic email domains sometimes filter transactional email.

In the Transkribus web interface, click New Collection. Give it a descriptive name (e.g., "Smith Family Papers - Correspondence" or "Board Minutes 1920-1950").

1. Click the + New Collection button
2. Enter a collection name that describes the documents
3. Click Create

What you should see: Your new collection appears in the list with 0 documents.

Click on your collection to open it. Click Upload Documents and drag your scanned image files into the upload window. You can upload individual images or group them as a multi-page document.

1. Click Upload Documents (cloud with up-arrow icon)
2. Click Select Files and choose your scanned JPG or TIFF files
3. Name the document (e.g., "Letter from John Smith to Mary Smith, March 1887")
4. Click Upload

What you should see: Progress bar during upload, then the document appears in your collection with a thumbnail of the first page.

Troubleshooting: If upload fails, check that files are JPG or TIFF format and under 50MB each. Very large files may need to be compressed first.

Open your document by clicking its thumbnail. You'll see the document image with an empty transcription panel on the right.

Click the Run HTR+ button in the toolbar (or look for "Automatic Text Recognition" in the Tools menu).

1. In the toolbar, click HTR+ (the lightning bolt icon with "HTR" label)
2. A model selection dialog appears. Select a pre-trained model appropriate for your document era:

   • For 19th–20th century English cursive: select "English Handwriting (1800-1900)" or "Transkribus HTR+"
   • For typed/printed documents: select an OCR model instead
3. Select which pages to process (all pages, or a range)
4. Click Run

What you should see: A progress indicator. Processing takes 30–90 seconds per page. When complete, text appears in the transcription panel on the right side of the screen.

Troubleshooting: If no text appears, the model may not match the document's script style. Try a different pre-trained model. Very degraded or extremely small handwriting may not transcribe well with any model.

The AI transcription will have errors, typically 5–15% of words for clear documents and more for degraded or unusual handwriting. Review the transcription panel against the image and correct errors.

1. Click on a word or line in the transcription panel
2. Compare it against the image; the corresponding area in the image is highlighted
3. Click to edit and type the correct word
4. Press Tab to jump to the next uncertain word (highlighted in yellow)

What you should see: Lines where the AI is uncertain are highlighted in yellow. The verification process goes much faster than transcribing from scratch.

When you've reviewed and corrected the transcription, export it in the format you need.

1. Click Export in the toolbar
2. Choose your format:

   • Plain text (.txt): for full-text search in your repository
   • DOCX: for editing in Word
   • PDF with text layer: for accessible PDF with hidden searchable text
   • TEI/XML: for scholarly publication or digital humanities projects
3. Click Download

What you should see: A file downloads to your computer containing the transcribed text.