Does MRZ scanning require an internet connection, or can it work on-device?

This depends entirely on how a given vendor has built their software, and it's a meaningful distinction to understand when evaluating a document scanning solution, since it affects privacy, latency, and reliability in different environments.

On-device (sometimes called on-premise or edge) MRZ scanning runs the entire extraction and validation process locally, on the phone or scanning device itself, without sending the document image to an external server. This has real advantages.

It works without an internet connection, which matters for use cases like airport gates or remote areas with unreliable connectivity. It's typically faster, since there's no network round-trip involved in getting a result.

And it can offer stronger privacy guarantees, since the sensitive document image and extracted personal data never leave the device unless the business chooses to transmit the final structured result afterward.

Cloud-based MRZ scanning, by contrast, sends the captured image to a remote server for processing and returns the extracted, validated data. This approach can leverage more computationally intensive processing since it isn't constrained by the device's own hardware, and it centralizes updates and improvements to the OCR engine without requiring an app update on every user's device.

The tradeoff is a dependency on network connectivity and, depending on the vendor's data handling policies, a different privacy posture, since the document image is transmitted somewhere outside the device.

Many identity verification vendors offer both options, or a hybrid where initial capture and basic checks happen on-device while more intensive processing (like deep authenticity analysis) happens server-side, giving businesses flexibility depending on their specific requirements around connectivity, privacy regulations, and processing needs.

ScanDoc's SDKs support this kind of flexible deployment, allowing businesses to choose the processing approach that fits their environment, a fully offline scanning flow for settings without reliable connectivity, or a cloud-connected approach where more intensive checks are handled server-side, rather than locking businesses into a single model regardless of their actual operating conditions.