Google Cloud Vision API Pricing: Google Vision API Pricing and OCR Cost Per 1,000 Pages

Google Cloud Vision API pricing at a glance

Google Cloud Vision charges per image, per feature. The first 1,000 units of each feature every month are free. After that, OCR through Text Detection or Document Text Detection costs $1.50 per 1,000 units up to 5 million units a month and $0.60 per 1,000 above 5 million. A PDF page counts as one image, so for document work a unit is effectively a page. There is no subscription, no minimum and no commitment tier.

All prices are USD per 1,000 units. One detail in that table matters for high-volume document work: the OCR features drop to $0.60 past 5 million units, a 60% cut, while Label Detection only drops to $1.00. If you are reading documents at scale, OCR is the feature with the steepest volume break Google offers on Vision.

How much does Google Vision API cost per page?

Google Vision OCR costs $0.0015 per page between 1,001 and 5 million pages a month, and $0.0006 per page beyond that, with the first 1,000 pages free every month. Google treats each page of a PDF or TIFF as a separate image, so a 3-page invoice is billed as 3 units for each feature you request, not as one document.

Pricing runs in 1,000-unit blocks and the last block is prorated. Google's own example: 5,300 landmark requests in a month cost $6.45, because the first 1,000 are free and the remaining 4,300 are 4.3 blocks at $1.50. The same math applies to OCR.

What Vision OCR costs at real invoice volumes

Here is the monthly Document Text Detection bill for common US accounts payable volumes, assuming one feature per page and the free 1,000 units applied first.

At 6 million pages the first 4,999,000 billable units cost $7,498.50 at $1.50 and the last 1,000,000 cost $600 at $0.60. Very few AP departments get near the 5 million break, so for practical budgeting the rate is $1.50 per 1,000 pages.

Is Google Cloud Vision API free?

Partly. Google Cloud Vision gives every feature 1,000 free units per month, permanently, so a team reading fewer than 1,000 pages a month with one feature pays nothing. Above that every unit is billed. New Google Cloud accounts also get $300 in trial credit, but you still need a billing-enabled project to call the API at all.

Text Detection vs Document Text Detection: which one to pay for

Both cost the same, so pick on output, not price. Text Detection is tuned for sparse text in photos, such as signs and labels. Document Text Detection is tuned for dense text and returns a page, block, paragraph, word and symbol hierarchy that is far more useful for invoices, statements and forms. For any invoice workload, Document Text Detection is the right feature, and calling both on the same page simply pays twice for the same words.

The limits that shape a Vision OCR invoice pipeline

The published quotas decide how you have to build the integration, and they are where most of the setup effort goes.

• Images: up to 20 MB per image and up to 16 images per synchronous images:annotate request.
• PDF and TIFF, synchronous: files:annotate reads at most 5 pages per file. If you do not name the pages, the quota counts 5 pages, though you are billed only for pages actually processed.
• PDF and TIFF, asynchronous: files:asyncBatchAnnotate accepts files up to 2,000 pages and 1 GB, reads input from Cloud Storage and writes JSON output files back to a Cloud Storage bucket.
• Authentication: API keys are not supported for asynchronous file requests, so batch PDF work needs a service account.

None of this is hard for an engineer who works in Google Cloud every day. It is, however, a Cloud Storage bucket, a service account, an operation poller and a JSON reader before you have a single invoice total.

Does Google Vision OCR extract invoice fields?

No. Google Cloud Vision returns the text on the page with bounding boxes and a reading hierarchy, but it has no concept of an invoice number, vendor, due date, tax or line item. To get those fields you either write parsing rules against the raw text, send the text to a language model, or pay for a separate extractor such as the Google Document AI Invoice Parser.

That is the real pricing decision. The table below puts Vision next to the options a US team usually compares when the goal is invoice data, not just text.

Two things stand out. First, Google charges exactly the same $1.50 per 1,000 for Vision OCR and for Document AI Enterprise OCR, and both give 1,000 free units a month, so the choice between them is about output format, not cost. Second, the jump from reading an invoice to understanding it is where Google's pricing climbs steeply: the Invoice Parser costs about 66 times the OCR rate on single-page invoices, because each document is billed as a full 10-page block.

Is Google Vision cheaper than AWS Textract and Azure for OCR?

For plain text they are priced identically. Google Vision Document Text Detection, AWS Textract Detect Document Text and Azure Document Intelligence Read all cost $1.50 per 1,000 pages at normal volumes and $0.60 at high volume. The difference is where the break happens: AWS and Azure drop to $0.60 after 1 million pages a month, while Google waits until 5 million. Between 1 and 5 million pages a month, Vision is the more expensive of the three.

So Vision only wins on price below 1 million pages if you value the 1,000 free units every month, which are worth $1.50. In practice, pick the cloud you already run on. Pick on field extraction if you need invoice data, because that is where the three vendors diverge by a factor of ten.

We run the invoice case for both clouds at 5,000 invoices a month in AWS Textract vs Google Vision pricing for invoice OCR.

When Google Vision is the right buy

Vision is the right buy when text is the product. Digitizing an archive for full-text search, feeding page text into your own model, reading labels in photos, or adding OCR to an app that already runs on Google Cloud all fit the per-unit model well, and $1.50 per 1,000 pages is a fair price for that.

It is the wrong buy when the real requirement is "get these vendor invoices into QuickBooks with every line item by Friday." In that case the OCR bill is the smallest number in the project and the parsing, review and export work is the largest. A finished invoice extractor reads the header and each line, lets a person check the result, and exports Excel, CSV or JSON at a flat monthly price. Upload one of your own invoices in the tool above and compare the output with a Vision OCR response before you decide which one your team should maintain.