Server-Side OCR: How to Process Documents at Scale Without Third-Party APIs

OCR is commonly used in server-side applications in a number of industries, for applications such as:

Insurance

Legal and Courts

Healthcare

Developers consider well-known OCR API services, such as those provided by Google, Amazon and Microsoft, as well as OCR SDKs. Each approach has benefits and tradeoffs. Cost and data control are key differences between SDKs and API services.

Apryse OCR offers a unique combination of high-speed performance, multilingual support, and high accuracy specifically designed for large enterprises and technology providers.

Sensitive data must be shared in compliance with regulations such as HIPAA and GDPR, depending on jurisdiction and industry-specific regulations.

API Services

Because data is leaving your environment and being processed by a third party, data security and privacy aredependent on third-party compliance with relevant standards.

Apryse SDK

With no external dependencies, your data never leaves your servers, making privacy and data security compliance simple.

SDKs like Apryse are licensed, while API services are typically pay-per-use. For example, Google Cloud Vision uses tiered pricing per thousand units. When it comes to cost, the optimal solution depends on scale. In addition, don’t forget to consider network data costs in your calculations.

API Services

While API services and their pay-as-you-go model are cheap at low volume, costs can quickly snowball out of control when OCR is needed at high volume. Because APIs like Google Cloud Vision scale linearly, sporadic and low-volume OCR may be cost effective with the API.

Apryse SDK

Apryse offers flexible licensing, there may be a license that works for your use case that allows your application to gain the benefits of Apryse without compromising on cost. making your costs more predictable. Depending on your license, an OCR SDK becomes more cost-effective than API services when:

To learn more about Apryse SDK pricing and licensing, please contact sales. Our team can help find the best approach for your needs.

Image pre-processing

Scanned documents aren’t always clean and aligned. For best performance and accuracy, pre-processing such as deskew and despeckle help prepare documents to get the best results.

With Apryse SDK, these pre-processing features are built into the engine, improving recognition accuracy.

In comparison, Google Cloud Vision, Amazon Textract and Microsoft Azure OCR don’t specifically mention deskewing or despeckling in any of their OCR documentation. Instead, these services suggest you send images after performing pre-processing separately. This workaround can get complicated, however, if you’re using batch processing or queue-based processing, and could also add latency.

Output formats

The Apryse OCR module supports output in structured JSON, XML and searchable PDF.

API services such as Google Cloud, Amazon Textract and Microsoft Azure OCR support JSON output, and rely on other tools of these platforms for other outputs, such as CSV or searchable PDF.

Architecture Considerations

Cloud OCR services are delivered as multi-tenant, microservice-based systems. Because requests traverse load balancers, authentication services, storage layers, async job managers, and other microservices, latency per document is variable. Community posts on some of these services reveal developers experiencing latency that varies widely, from hundreds of milliseconds to over 5 seconds per document. This variability is unpredictable, and can cause workflow bottlenecks.

In comparison, because Apryse SDK runs on your own services, the OCR engine performs in-process without this added, variable latency.

When it comes to getting OCR jobs done, the comparison is a lot like the difference between waiting for an uber or owning your own car.

With Apryse SDK, you choose batch processing or queued processing based on your needs, not based on compromise.

To find out more on how to set up the Apryse Server SDK with the OCR module, visit our documentation.

To get started with OCR, you’ll first need the Apryse Server SDK. The Server SDK supports a range of frameworks, runtimes and languages in all major platforms (MacOS, Windows, Linux), so you can deliver applications from a single codebase.

Frameworks and runtimes include:

Supported languages:

If you don’t see the language you require, contact sales for more information.

For this article, we’ll use Node.js. You can follow this tutorial video on How to Use Apryse SDK in Node.js on Windows.

Before you start:

Once you have the SDK installed, you can follow the instructions on our Node.js Getting Started documentation page to download and install a sample package that demonstrates our PDF conversion capability to test your setup. You’ll also follow the instructions to enter your license key into the module.

With these steps, you can learn how to set up your environment, import the required libraries, and run a script that produces a valid PDF file, all without manual intervention. This example provides a practical foundation for more advanced document‑generation workflows.

This creates a new package.json file:

At the command prompt, type exit, press Enter, and close the window. 

Open Visual Studio Code, select File >Open Folder. Navigate to Document > NewApryseProject and click on Select folder.

Next, from the File menu, select NewTextFile... and name it index.js. Then copy the following code and paste into the index.js file to create a blank PDF page.

Keep in mind the following about the code above:

Next, add your license key. Scroll down to the line containing PDFNet.runWithCleanup(main, ‘YOUR_PDFTRON_LICENSE_KEY’) and replace the words in quotes with the copy of your trial license key.

After you save your changes and close the file, you have successfully added the Apryse Server SDK to your Node.js project. Now you’re ready to get started with the OCR module!

To run your project, navigate to the Documents folder. Select the NewApryseProject folder, right-click and select Open in Terminal.

From the project directory, on the command line, enter the following, then press Enter:

The sample code executes, creates a blank PDF page, and displays the message, “Hello World!”

Now, exit out of the command line and navigate to the Documents > NewApryseProject folder. Open the blank.pdf which you created by integrating the Apryse Server SDK!

To get started with the OCR module, first download the module for Windows on our add-on modules documentation page.

Add-on modules for Server/Desktop library

Unzip the file and install the package.

Here’s what you can do with the OCR module. If only one module is present (IRIS module or default OCR module), Apryse SDK will use that module. If both are present, you can select the module you wish to use with the OCR options object:

OCROptions.setEngine("iris")

Find a full code sample for converting images to PDF with searchable text here.

To make a searchable PDF by adding invisible text to an image based PDF such as a scanned document using OCR.

For more workflow examples, sample code and instructions, check out the OCR Workflow documentation page.

Q: What is the difference between SDK and API?

A: A software development kit (SDK) puts a complete set of tools including libraries, documentation, and utilities. In comparison, an application programming interface (API) refers to a set of rules that allow your software to communicate with an external software. Check out our article, Local SDKs vs. Cloud APIs: What Developers Need to Know Before Building the Next Application to learn more.

Q: Is OCR the best data extraction option?

A: Not always. While OCR is great for turning human-readable text (images, PDF, scans) into machine-readable text, it doesn’t always capture document structure or other context data. Check out Smart Data Extraction for an AI-powered solution that gets more data out of your docs.

Q: Can I set up OCR in Python?

A: Yes, Apryse SDK supports a number of languages for the OCR SDK, including Python. Check out How to Build Optical Character Recognition (OCR) in Python for instructions.

Unleashing The Power of OCR with JavaScript

Read Barcodes and Convert to Searchable PDF with C++ using Apryse SDK

Streamline Document Workflows with Apryse’s OCR and Barcode Extraction