Apryse OCR vs Tesseract vs Cloud OCR API: Tested in Python

Apryse OCR vs Tesseract is a choice between a commercial on-premise SDK and the best-known open-source engine. There is a third option: a cloud OCR API that you call over HTTP. We installed all three in Python and ran them on the same four real documents, then counted how many key facts each one read correctly. Every number below comes from that test.

The Short Answer

• Tesseract: free, runs anywhere, good on clean prints. On worn receipts and old scans it needs preprocessing and tuning.
• Apryse OCR: an add-on to the Apryse Server SDK. It runs on your servers, turns scans into searchable PDFs, and fits teams that already use Apryse for PDF work. Pricing is by quote.
• Cloud OCR API: no install and no model files. In our test, the OCR Wizard API read every key fact on all four documents.

The Test Documents

We picked documents that look like real work, not clean screenshots. All are free to reuse (see Sources):

1. A 1994 Tesco receipt on thermal paper, faded and creased.
2. A 2020 Polish supermarket receipt, photographed with a phone.
3. Two typewritten receipts from 1913, scanned by the U.S. National Archives.

For each document we listed the facts a business would need: prices, totals, dates, names and amounts. That gave 33 facts in total. An engine scores a point when the fact appears exactly right in its output.

Results: Key Facts Read Correctly

Tesseract ran with page segmentation mode 6, which scored better than the default mode 3 (9 facts). Apryse ran with its two documented profiles: default (fast) and accurate (slower). The accurate profile helped a lot on the 1913 scans.

Here is the same part of the 1913 rental receipt, as read by each engine:

On this scan, Tesseract turned $25.00 into $2b.00 and $12.50 into $12,600. Apryse dropped letters from the written amount and returned $350. 00. with a space, which breaks a simple amount parser. The API returned the three lines as printed.

To be fair to both local engines: on a clean, sharp image of printed text that we generated, Apryse and Tesseract both read every word. Apryse also says its module is optimized for color or greyscale input at about 300 DPI. The gap appears on faded, low-contrast and old documents, which is exactly what receipt and archive projects deal with.

Speed

Time per page on the same documents (local engines on a laptop CPU, API including the network):

• Tesseract 5.3: 0.3 to 1.2 seconds.
• Apryse, default profile: 1.7 to 4.5 seconds.
• Apryse, accurate profile: 6 to 15 seconds.
• OCR Wizard API: 0.6 to 0.8 seconds.

The Code for Each Engine

Tesseract

Install the engine (apt install tesseract-ocr on Debian or Ubuntu), the language data you need, and pip install pytesseract:

import pytesseract

text = pytesseract.image_to_string("receipt.jpg", lang="eng", config="--psm 6")
print(text)

Apryse

Install the SDK and the OCR module from the Apryse package index, then set your trial or license key in APRYSE_KEY:

pip install --extra-index-url=https://pypi.apryse.com apryse-sdk apryse-ocr

import json
import os

import apryse_ocr
from apryse_sdk import PDFDoc, PDFNet, OCRModule, OCROptions

PDFNet.Initialize(os.environ["APRYSE_KEY"])
PDFNet.AddResourceSearchPath(os.path.join(os.path.dirname(apryse_ocr.__file__), "lib"))

doc = PDFDoc()
options = OCROptions()
options.AddLang("en")
OCRModule.ImageToPDF(doc, "receipt.jpg", options)
doc.Save("receipt_searchable.pdf", 0)

data = json.loads(OCRModule.GetOCRJsonFromPDF(doc, None))
for page in data["Page"]:
    for para in page["Para"]:
        for line in para["Line"]:
            print(" ".join(word["text"] for word in line["Word"]))

Two details cost us time. The SDK only finds the OCR module after AddResourceSearchPath points to the module's lib folder. And the module expects two-letter language codes such as en and pl: the three-letter code pol made the call fail.

OCR Wizard API

Only requests and a RapidAPI key in RAPIDAPI_KEY. The language is detected automatically:

import os

import requests

with open("receipt.jpg", "rb") as f:
    response = requests.post(
        "https://ocr-wizard.p.rapidapi.com/ocr",
        headers={
            "x-rapidapi-host": "ocr-wizard.p.rapidapi.com",
            "x-rapidapi-key": os.environ["RAPIDAPI_KEY"],
        },
        files={"image": f},
        timeout=60,
    )
response.raise_for_status()
result = response.json()["body"]
print(result["detectedLanguage"])
print(result["fullText"])

The response also lists every word with its bounding box. Our document OCR tutorial uses those boxes to pair receipt items with their prices.

Setup, Licensing and Deployment

Which One Should You Pick?

• Pick Tesseract for clean prints, offline tools and hobby projects, when you have time to tune preprocessing and page modes.
• Pick Apryse when documents must stay on your own servers and you already use the Apryse SDK to view, edit or convert PDFs. Test the accurate profile on your own scans, and ask their sales team for a quote early.
• Pick a cloud OCR API when your inputs are photos, receipts or old scans, when you want results in a few lines of code, and when you would rather pay per request than run OCR servers.

For more open-source options, including EasyOCR and PaddleOCR, see our comparison of OCR APIs and open-source engines. For PDFs, the PDF OCR in 5 lines guide shows the API's PDF endpoint.

How We Tested

• Tesseract 5.3.0 with the official English and Polish language data, through pytesseract.
• Apryse SDK 12.2 and the apryse-ocr 12.2.1 package from the Apryse package index, run with a free trial key, languages en and pl, in both profiles.
• OCR Wizard API, default settings, the same image files.
• A fact counts only when it appears exactly as printed, for example $393.75.

Next Step

Run the three snippets on one of your own documents. To try the API, get a key on the OCR Wizard API page.

Sources

• Apryse OCR guide (modules, languages, IRIS)
• Apryse OCR for documents (licensing, input quality, add-ons)
• Tesseract on GitHub (Apache 2.0)
• Tesco grocery receipt, Finchley 1994 (Wikimedia Commons, public domain)
• Polish supermarket receipt by Hippietrail (Wikimedia Commons, CC BY-SA 4.0)
• Receipt for rental of Overland Park, 1913 (U.S. National Archives, public domain)
• Receipt for payment of watchman, 1913 (U.S. National Archives, public domain)