Guides and pricing comparisons on OCR, AI document data extraction, and intelligent document processing for teams evaluating tools.
A Document Transaction is the unit Adobe bills PDF Services API in. Extract gets 5 pages per transaction, OCR gets 50, and Accessibility Auto-Tag burns 10 per single page. Here is exactly how the counting works.
Read articleKofax Capture and Kofax TotalAgility are two different products that are often confused for versions of each other. Here is what each one actually does, which one you need, and what the Tungsten rename changed about both.
Read articleAmazon Textract, Azure Document Intelligence and Google Document AI all read handwriting, and all three bill it at the same rate as printed text. The real difference is language coverage, which collapses from 300+ printed languages to 12 on Azure and from six to one on Textract. Here is the comparison from the vendors' own documentation.
Read articleOpenText sells two capture products with confusingly similar names. Core Capture is the cloud SaaS line built on continuous machine learning. Intelligent Capture is the deployed enterprise platform that used to be called Captiva. Here is how to tell which one you are running, and which one fits.
Read articleAWS and Azure halve their basic OCR rate above one million pages a month. Google does not until five million. All three list at the same $1.50, which makes a list price comparison the wrong tool at volume.
Read articleZDR covers the stateless inference call. A document pipeline is the one workload that is not stateless, and the file upload step is named in every vendor's exclusion list. The gap between having ZDR and running under it.
Read articleAWS may store customer content for service improvement, including model training, and may store it in a different region. The opt-out exists, it is real, and it is switched off by default. Here is what the documentation actually says.
Read articleThe model is free and the GPU is $587.50 a month whether it reads a page or not. Real AWS rates, the effective cost per 1,000 pages at every volume, and why the widely quoted $176 per million figure describes a saturated fleet rather than your workload.
Read articleOpenAI counts images in 32 by 32 pixel patches, shrinks anything over the model budget, then applies a per-model multiplier. A scanned letter page is 2,080 patches on a full model and 1,496 on mini or nano. Here is the full calculation.
Read articleYes, but only Microsoft ships containers among the three hyperscalers, and even those phone home to Azure for billing. Here is what actually runs on your own hardware in 2026, what it costs, and how to tell whether your policy really requires it.
Read article