ParseFlow: The OCR API That Extracts Forms & Tables for 88% Less Than Textract

Sep 10, 2026 · Send a PDF or image to one endpoint and get back clean JSON with text, forms, and tables — at a fraction of AWS Textract's price. Per-page billing, Textract-compatible responses, and a $5 free trial.

Extracting structured data from documents is a classic engineering trap. The moment you outgrow copy-pasting, you're looking at AWS Textract or Google Document AI — and then the "simple" integration turns into a project. S3 buckets, IAM roles, Lambda functions, raw block graphs you have to reassemble, and a bill that surprises you again every month.

ParseFlow exists to make document extraction feel like it should: one POST request, structured JSON back. And it prices itself like a tool built by someone who used to pay Textract's invoices.

What is ParseFlow?

ParseFlow is a structured OCR API — a high-performance, cost-effective alternative to AWS Textract and Google Document AI. Send any image or PDF, get back clean JSON with text, form fields, and tables, typically within seconds.

There's no infrastructure to run and nothing to provision:

curl -X POST https://api.parseflow.cc/v1/document/analyze \
  -H "Authorization: Bearer ocrg_live_key" \
  -F "document=@invoice.pdf" \
  -F 'featureTypes=["FORMS","TABLES"]'

Response (native format):

{
  "DocumentMetadata": { "Pages": 1 },
  "Text": "Invoice #1042\nBilled to: Acme Corp...",
  "Forms": [
    { "key": "Invoice Number", "value": "1042", "confidence": 0.98 }
  ],
  "Tables": [{ "rows": [["Item", "Qty", "Price"]] }],
  "Confidence": 0.97,
  "Billing": { "totalCostCents": 1, "pages": 1 }
}

Inputs are PDF (up to 40 MB, 50 pages), PNG, and JPEG/JPG. Send it as a multipart file upload or as base64 bytes in a JSON body.

How it works

Behind a single endpoint, ParseFlow handles the plumbing: layout parsing, confidence checking, model escalation, and parallel PDF processing.

Text, forms & tables — scanned images and PDFs come back with plain text, key-value form fields, and table rows already parsed — no manual mapping required.
Multi-page PDFs — documents are auto-detected, split page by page, and processed in parallel, with per-page billing and confidence scoring.
Confidence-based review — low-confidence pages are automatically retried on a larger vision model, and flagged for human review if still uncertain. The escalation surcharge applies only when it's needed, and only to that page.

Feature types and per-page pricing

The featureTypes array controls which extraction features run, billed at the rate of the most expensive tier requested:

Feature	Price / 1K pages	Extracts
`TEXT` (default)	$1.00	Plain text in reading order
`TABLES`	$5.00	Table detection and cell extraction
`EXPENSE`	$4.00	Receipt / expense-specific fields
`FORMS`	$8.00	Key-value field pairs with per-field confidence

ParseFlow vs AWS Textract

Feature	ParseFlow	Textract	You save
Text extraction	$1.00	$1.50	33%
Table extraction	$5.00	$15.00	67%
Form/field extraction	$8.00	$50.00	84%
Forms + Tables (combined)	$8.00	$65.00	88%
Expense analysis	$4.00	$10.00	60%

For forms + tables at scale: 1,000 pages is $8 on ParseFlow vs $65 on Textract — a saving of $57. At 10,000 pages it's $80 vs $650 ($570 saved), and at 100,000 pages, $800 vs $6,500 ($5,700 saved).

And it's not only price. Textract's forms+tables pipeline requires S3, IAM, and Lambda setup, bills failed attempts, returns a raw block graph rather than structured JSON, and multi-page documents cost 10–20× the single-page rate. ParseFlow requires a single API key, bills only successful pages, and returns structured JSON out of the box.

A drop-in replacement for your existing integration

You don't have to rewrite code to switch. ParseFlow is designed as a drop-in replacement for both Textract and Document AI, switchable from Dashboard → Settings → Response format:

AWS Textract-compatible — returns Blocks[] with PAGE, LINE, KEY_VALUE_SET, TABLE, and CELL block types. Swap the endpoint to https://api.parseflow.cc/v1/document/analyze, replace your AWS credentials with the Bearer token, and your existing code that parses Blocks[] works unchanged.
Google Document AI-compatible — returns document.text, pages[].formFields, entities[], and documentLayout.blocks, so existing DocAI parsers keep working.

No S3 buckets, IAM roles, Lambda functions, or complex infrastructure setup.

Billing that's honest about complexity

Prepaid wallet model — credits are deducted after each successful extraction. You get $5.00 of free trial credits on first signup, and top up from the dashboard via Razorpay Checkout.
Per-page, no subscriptions — pay per page processed, no minimums, no monthly commitment. You can revoke API keys at any time and cancel whenever you like.
Only successful pages are billed — pages that fail to render come back with an Error field and are not charged.
Usage dashboard — full transaction history, invoices, and settings, with payment links delivered by email or SMS.

Rate limits are 60 requests per minute per IP address (HTTP 429 with a Retry-After header on overflow). Typical latency: 300–800 ms for a single image, ~3–8 seconds for a 10-page PDF, and ~15–30 seconds for a full 50-page PDF.

What people use ParseFlow for

Invoice OCR — extract invoice numbers, line items, dates, and totals into structured fields.
Receipt OCR — expense-specific fields for receipts, including the built-in EXPENSE feature type.
Table extraction — turn scanned tables into row/column data ready for a spreadsheet or database.
Form recognition — key-value fields for applications, registrations, and paperwork.
Bank statement & identity documents — statements, passports, and driving licenses (TIFF support available on request).

Scanned PDFs work with the same vision pipeline; for handwriting support it depends on the underlying model — typed and printed text is preferred. Password-protected PDFs aren't supported, so remove protection before uploading.

Frequently asked questions

What file types does ParseFlow support?

PDF, PNG, and JPG/JPEG, up to 40 MB per upload and 50 pages per request. TIFF support is available on request. Larger documents should be split client-side.

How do I migrate from AWS Textract?

Generate a ParseFlow API key, set your response format to "AWS Textract-compatible" in dashboard settings, replace your endpoint URL with https://api.parseflow.cc/v1/document/analyze, and swap your AWS credentials for the Bearer token. Your existing code that parses Blocks[] works unchanged.

How is billing calculated?

Per page processed, at the rate of the most expensive feature tier you request (e.g., FORMS is $8 per 1,000 pages). Only successful pages are billed, and a page that needs escalation to the more accurate model is charged the surcharge only for that page. There are no subscriptions and no minimums.

What response formats are available?

Three: the ParseFlow native format (structured Text/Forms/Tables/Pages/Billing), an AWS Textract-compatible Blocks[] format, and a Google Document AI-compatible format. You switch between them in dashboard settings — no API changes needed.

What's the free trial?

New signups get $5.00 in wallet credits on first signup — enough to process hundreds of pages before you pay anything. There's no credit card required to start.

Get your API key in 30 seconds

No credit card required to start — you get $5.00 in free trial credits on signup, and top up as you go. If you're migrating off Textract or Document AI, the compatibility formats mean your existing code keeps working.

Create your free account and get an API key → Start with ParseFlow