Plain text OCR costs about $1.50 per 1,000 pages on AWS Textract, Google Document AI, and Azure AI Document Intelligence. Structured receipt and invoice extraction costs far more: Textract AnalyzeExpense is $0.01 per page, Azure prebuilt models are $10 per 1,000 pages, and Google charges $0.10 per document of up to 10 pages, which makes a single-page receipt ten times more expensive than a ten-page one. Every figure on this page was read off the vendor pricing page or official price API in August 2026. Upload a receipt below to see the same structured output without wiring up an API first.
Upload your receipts and invoices
Drop files here or click to upload
Up to 50 files
Uploading...
Every vendor prices a different unit. One charges per page, one per document of up to ten pages, one per credit, and one per workflow step. Two APIs can look identical on a price list and differ by 10x on the same stack of receipts, so the only honest comparison is the one that runs your actual document mix through each billing model.
AWS and Azure bill per page. Google Document AI bills per document, counting every ten pages as one unit. Mindee bills per credit, where one credit equals one physical page. Nanonets bills per workflow block executed. Comparing the headline numbers without converting to a common unit gives you the wrong answer every time.
The cheap number everyone quotes is plain text detection. The moment you ask for tables, forms, key-value pairs, or queries, the rate stacks. On Textract, a page with forms, tables, and queries together is $0.070, which is roughly 47 times the $0.0015 raw OCR rate on the same page.
A deployed Google custom processor bills $0.05 per hour of hosting whether or not you send it a document, which is about $438 a year for a processor left running. Azure custom neural training is free for the first 10 hours and then $3 per hour. Failed and retried calls, storage, and the engineering time to normalize raw JSON into a usable row never appear on any pricing page.
The AWS free tier lasts three months and covers only 100 AnalyzeExpense pages a month. Azure free tier resources analyze only the first two pages of any document, so a five-page invoice comes back incomplete. Free tiers are useful for a proof of concept and misleading as a basis for a budget.
Below is the list price for structured receipt and invoice extraction on every major document API, normalized to the same unit and checked against the vendor source. Where a vendor publishes no flat price, that is stated instead of guessed. Nothing here is estimated or scraped from a review site.
Amazon prices its receipt and invoice endpoint at $0.01 per page in US West (Oregon) for the first million pages a month, dropping to $0.008 per page after that. OCR output is included in the response at no extra charge.
The Google expense parser (formerly the receipt parser) and invoice parser both cost $0.10 per document, where one count covers up to 10 pages. A one-page receipt costs $0.10; a ten-page invoice also costs $0.10.
Azure AI Document Intelligence prices its prebuilt receipt and invoice models at $10.00 per 1,000 pages on the S0 tier in East US, which works out to $0.01 per page, matching Textract exactly.
If you only need text, Textract DetectDocumentText and Azure Read are both $1.50 per 1,000 pages, and Google Enterprise Document OCR is free for the first 1,000 pages a month then $1.50 per 1,000. You still have to write the parsing yourself.
Google custom extractor and Azure custom extraction are both $30.00 per 1,000 pages, three times the prebuilt rate, before training and hosting. Textract Custom Queries is $0.025 per page with no free tier at all.
The API fee is usually the smallest line. Normalizing raw JSON, handling multi-page splits, retrying low-confidence pages, and building an export are engineering hours that no per-page rate includes.
Work from your real document mix, not the headline rate. Most teams overestimate the API line and underestimate everything around it.
Pull a month of real files and count physical pages. AWS, Azure, and Mindee all bill per page, so page count is the base unit for three of the four pricing models you will compare.
Tip: Average pages per document matters enormously for Google, where a 1-page receipt and a 10-page invoice cost the same $0.10.
If you need only the text, price the raw OCR endpoint at about $1.50 per 1,000 pages. If you need vendor, date, tax, totals, and line items as fields, you are pricing the expense or invoice model, which is roughly 7x that.
Budget for the pages you send twice because confidence was low. Add $0.05 per hour if a Google custom processor stays deployed, and $3 per hour of Azure custom neural training past the free 10 hours.
Put a number on the developer weeks needed to normalize output, split multi-page files, and build the export. Compare that total against a flat monthly tool, then pick the cheaper path for your volume.
Built for the person who has to defend a number in a budget: the engineer sizing a build, the finance lead sizing a buy, and the firm deciding which one to bill a client for.
You have been asked what receipt extraction will cost at 50,000 pages a month and need a defensible number that includes retries and normalization, not just the per-page rate.
You are comparing a per-page cloud API against a flat monthly tool and need the crossover volume where one stops being cheaper than the other.
You process client receipts at volume and need to know whether per-page API billing or a seat-free flat plan gives you a better margin per client.
You want to know whether wiring up Textract yourself beats paying for a finished extraction product once engineering time is counted honestly.
Last updated: August 2026. Every rate below was read from the vendor pricing page or official price API on 2026-08-03.
List prices in US dollars, standard pay-as-you-go tier, US regions (Textract figures are US West Oregon, Azure figures are East US S0). Committed-use and enterprise contracts are lower and are not shown, because vendors do not publish them.
| API | Receipt / invoice extraction | Plain text OCR | Custom model | Free tier |
|---|---|---|---|---|
| Amazon Textract | AnalyzeExpense $0.01 per page ($10 per 1,000), $0.008 above 1M pages/mo | DetectDocumentText $0.0015 per page ($1.50 per 1,000), $0.0006 above 1M | Custom Queries $0.025 per page | 100 AnalyzeExpense pages/mo for 3 months |
| Google Document AI | Expense parser and invoice parser $0.10 per document, where 1 count covers up to 10 pages | Enterprise Document OCR $1.50 per 1,000, $0.60 per 1,000 above 5M | Custom extractor $30 per 1,000 pages, plus $0.05 per hour hosting | First 1,000 OCR pages per month |
| Azure AI Document Intelligence | Prebuilt models $10 per 1,000 pages ($0.01 per page) | Read $1.50 per 1,000 pages | Custom extraction $30 per 1,000 pages, neural training free for 10 hours then $3/hour | F0 tier, but it reads only the first 2 pages of any document |
| Veryfi | $0.08 per receipt, $0.16 per invoice, $0.25 per bank statement | Not sold separately | Enterprise only | 100 documents per month |
| Mindee | Starter from $44/mo, Pro from $116/mo; 1 credit = 1 physical page; extra credits from $0.05 | Included in the credit | Included on paid plans | 14-day trial |
| Nanonets | Per workflow block: $0.02 simple, $0.10 standard AI, $0.30 complex AI; a typical invoice flow runs 4 to 6 blocks | Included in the block | Included, no seat licenses | $50 in credits |
| ReceiptOCR | Starter $49/mo for 2,500 pages (about $0.020 per page), Plus $149/mo for 10,000 pages (about $0.015 per page); annual billing roughly halves both | Included | Custom extraction templates on Plus | Free to try, no card |
A plain text OCR API costs about $0.0015 per page, or $1.50 per 1,000 pages, on AWS Textract and Azure. Structured receipt and invoice extraction costs about $0.01 per page on Textract and Azure, and $0.10 per document on Google Document AI. Custom trained models cost around $0.03 per page. In other words, asking for fields instead of text multiplies the bill by roughly seven.
Google is the outlier, and the difference is easy to miss. Its expense and invoice parsers charge $0.10 per document, and one billed count covers up to 10 pages. If your documents are ten-page invoices, that is $0.01 per page and Google is competitive. If your documents are single-page coffee receipts, that is $0.10 per page, ten times what Textract and Azure charge for the same job. Average pages per document is therefore the single biggest driver of which API wins on price for you, and it is the number most comparison articles never mention.
We should say this plainly, because a comparison that only flatters the author is not worth reading. If you need raw text off a page and nothing else, Textract DetectDocumentText at $1.50 per 1,000 pages is roughly thirteen times cheaper than our effective per-page cost, and you should use it. If you are extracting millions of pages a month and already have an engineering team maintaining the pipeline, per-page cloud pricing with volume discounts will beat any flat plan. Our receipt OCR API makes sense in the middle: you want vendor, date, tax, line items, and totals as clean fields, at a volume in the thousands rather than the millions, without paying an engineer for a month to build the normalization layer.
For plain text, Google Enterprise Document OCR is cheapest because the first 1,000 pages each month are free, then $1.50 per 1,000. For structured receipts and invoices, Amazon Textract AnalyzeExpense and Azure prebuilt models tie at $0.01 per page. For very low volume, Veryfi is effectively free at 100 documents a month. The cheapest option changes with volume, so run your own page count through each row of the table above.
Volume discounts on the big three start higher than most teams reach. Textract drops from $0.0015 to $0.0006 per page for raw OCR and from $0.01 to $0.008 for AnalyzeExpense only after one million pages in a month. Google drops Enterprise Document OCR from $1.50 to $0.60 per 1,000 only above five million counts, and drops the custom extractor from $30 to $20 per 1,000 above one million. Azure sells committed tiers instead, where you prepay a monthly block and pay a lower overage rate. Below about 100,000 pages a month, assume you are paying list price everywhere.
Three line items regularly double a projected OCR budget. The first is idle hosting: a Google custom processor bills $0.05 per hour while deployed, whether or not it processes anything, which is about $438 over a year. The second is training: Azure gives you 10 free hours of custom neural training on v4.0 and then charges $3 per hour. The third, and by far the largest, is engineering. A raw API response is a bag of key-value pairs with confidence scores, not a spreadsheet row. Someone has to map fields, split multi-page files, decide what happens when confidence drops below threshold, and build the export. Two developer weeks at a loaded US rate costs more than 500,000 pages of Textract AnalyzeExpense.
Generally no. AWS does not bill for requests that return 4xx or 5xx errors, and Google states the same for failed requests. What you do pay for is a successful call that returns a low-confidence result you then send again, which is billed twice. Budget a few percent of your page volume for reprocessing, and more if you are sending crumpled thermal receipts or phone photos taken at an angle.
Run the math at your real volume. At 2,500 pages a month, Textract AnalyzeExpense costs $25 in API fees, which looks unbeatable next to a $49 plan until you add the pipeline someone has to build and keep running. At 500,000 pages a month, the API fee is $5,000 and a flat plan stops making sense, so building is clearly right. The crossover for most US teams lands somewhere in the tens of thousands of pages a month, and it moves depending on whether you already employ someone who can own the integration. Our cost comparison against manual data entry covers the third option, which is still what most small firms actually do.
For receipts, use an endpoint trained on receipts rather than a general OCR call. Textract AnalyzeExpense, the Google expense parser, and the Azure prebuilt receipt model all return merchant, date, total, and tax as named fields, which is the work you would otherwise write yourself. Our comparison of Google Vision OCR against Document AI covers why the cheaper general OCR endpoint is usually a false economy for receipts, and the Textract alternative page covers the same trade-off on the AWS side.
Vendors reprice, rename products, and move pricing behind sales forms. Google renamed its receipt parser to the expense parser. Azure renamed Form Recognizer to Document Intelligence. Klippa took its public pricing page down entirely, and Dext has never published a flat rate, so any article quoting a specific Dext or Klippa number invented it. Every figure on this page carries the date it was checked, and figures we could not verify from a vendor source are marked as not published rather than filled in with a plausible guess.
Plain text OCR costs about $1.50 per 1,000 pages on AWS Textract and Azure AI Document Intelligence, and Google gives you the first 1,000 pages free each month before charging the same $1.50. Structured receipt and invoice extraction costs about $10 per 1,000 pages on Textract and Azure, and $0.10 per document on Google Document AI. Custom trained models run around $30 per 1,000 pages.
Amazon Textract costs $0.0015 per page for DetectDocumentText, $0.01 per page for AnalyzeExpense on receipts and invoices, $0.015 for tables, $0.05 for forms, and $0.070 for forms, tables, and queries together, in US West (Oregon) for the first million pages a month. Rates drop above one million pages, for example AnalyzeExpense falls to $0.008 per page.
Google Document AI charges $0.10 per document for the invoice parser and the expense parser, where one billed count covers up to 10 pages. Enterprise Document OCR is free for the first 1,000 pages a month, then $1.50 per 1,000. Custom extractors and the form parser are $30 per 1,000 pages, plus $0.05 per hour to keep a custom processor deployed.
Azure AI Document Intelligence on the S0 tier costs $1.50 per 1,000 pages for Read, $10 per 1,000 pages for prebuilt models including receipt and invoice, $30 per 1,000 pages for custom extraction, and $3 per 1,000 pages for document classification. Custom neural training is free for the first 10 hours on v4.0 and $3 per hour after that.
For plain text extraction, Google Enterprise Document OCR is cheapest because the first 1,000 pages each month cost nothing. For structured receipt and invoice data, Amazon Textract AnalyzeExpense and Azure prebuilt models tie at $0.01 per page. The genuinely cheapest option depends on your page count and how many fields you need, so convert every vendor to a cost per 1,000 pages before deciding.
Only briefly. The AWS free tier lasts three months for new customers and covers 1,000 DetectDocumentText pages a month, 100 AnalyzeExpense pages a month, 100 AnalyzeID pages, 2,000 AnalyzeLending pages, and 100 pages of forms, tables, or queries. There is no free tier for Custom Queries. After three months you pay list price on every page.
It depends on the vendor, and the difference is large. AWS Textract, Azure, and Mindee bill per physical page. Google Document AI bills per document and counts every 10 pages as one unit, so a single-page receipt and a ten-page invoice both cost $0.10. Nanonets bills per workflow block executed. Always convert to a common unit before comparing.
Below roughly ten thousand pages a month, buying is usually cheaper once engineering time is counted, because the API fee is small next to the developer weeks needed to normalize output and maintain the pipeline. Above a few hundred thousand pages a month, per-page cloud pricing with volume discounts wins clearly. The API fee itself is rarely the deciding factor at small volume.
Stop typing receipts by hand
Upload your receipts and invoices and get a clean Excel or CSV file in minutes.
Extract my receipts nowFree to try, no sign up required
Send a receipt, get structured JSON back, on flat monthly pricing.
The invoice endpoint, line items and totals included.
Where AnalyzeExpense wins, and where a finished tool wins.
Why per-document billing punishes single-page receipts.
Prebuilt models compared with a no-code extraction tool.
