New release: November 2026

On-premise OCR server: four steps from test to production

Banks, insurers and mid-sized firms keep their pages in the building, so the OCR comes to the pages and not the pages to the OCR. This is the path from the first test page to production, on your own hardware or on a managed server in Germany: on-premise, offline after the install, with the data processing agreement ready to read. Tests need no application: you download, check, start.

On-premise OCR deployment: four steps from test to production

  1. 01

    Test the OCR on your worst pages

    Run the pages your last tool got wrong through the package on your own machine. No GPU yet? Test one page free, or ask for a test with more pages. Every word comes back with its box and a confidence, so you judge the result yourself. Your security team gets the published findings: 24 weaknesses, found and fixed.

    See real results Security findings

  2. 02

    Terms, data processing agreement and written offer

    Terms, processing agreement and technical measures are ready to read while you test. Price, term and delivery go into one written offer.

    Read the terms

  3. 03

    One package, one installer, one SHA-256

    One page tells your IT lead what the server needs and which graphics card carries which pipeline. You compare one SHA-256, run the installer, fetch the models. If an update fails, one command takes you back.

    System requirements Install and roll back

  4. 04

    Connect the OCR API and go live

    Your developers send a job, wait and fetch the result; the API comes with an OpenAPI file. Every pipeline has its pages per hour, with the run named, so you plan with numbers.

    API guide Pipelines and speed per GPU

On-premise OCR server operations: backup, package updates, support

Once it runs, running is all that is left.

  • Monitoring, hourly backup and restore

    On a managed server, alarms reach the operator, the backup runs every hour and the restore is tested. A release that fails its check rolls itself back.

    Update and roll back

  • Updates with pinned pipeline versions

    A pipeline has a name and a version, and a version stays as released. Releases go forward, and a failed update rolls back.

    Pipelines and versions

  • Support and maintenance

    Maintenance and repairs are part of a managed server. Support hours and response times go into the written offer. Your results are standard formats, so you can always take them elsewhere.

    Questions and answers

OCR server price, term and delivery

One written offer holds them, with real numbers instead of a list price.

Production use, high volume or your own server?

Private use and tests are free. For a company in production, a server of your own or a pipeline built on your pages, send the kind of page and the volume, and get a written offer, or a clear no when the fit is not there.

Ask for an offer