Pay as You Go
No commitment
£15 $15 €15 / 100 pages
One-time purchase. Valid for 1 year.
- AI-enhanced formatting
- Export to Markdown (plain text)
- Export to Microsoft Word
- Two-factor authentication
- API access
- No commitment
- Valid for 1 year
AWS Textract alternative
AWS Textract was built for printed forms. This was built for handwriting.
AWS Textract is excellent on printed invoices and forms at scale, but it slips on real handwriting. Handwriting OCR reads cursive and messy notes cleanly, suggests the fields to extract, and runs from a simple API on every plan.
- ✓Higher accuracy on cursive, messy and historical hands, not just print
- ✓Pulls handwritten forms into structured fields, schema suggested for you
- ✓A first-party API on every plan, with no S3, IAM or AWS setup
- ✓Private by default: encrypted, never used to train AI, auto-deleted
Complete
AWS Textract vs Handwriting OCR
Printed documents at scale, or handwriting done right
AWS Textract is a powerful cloud service for printed forms, tables and invoices, built into the AWS stack. Handwriting OCR is built to read handwriting and hand back structured data you can use straight away, with nothing to wire up.
Feature
AWS Textract
Handwriting OCR
Setup
AWS account, S3, IAM, SDK
Upload and go, or a simple API call
Built for
Printed forms, tables, invoices at scale
Handwriting: cursive, messy and historical
Handwriting accuracy
English print and handwriting, weaker on cursive
Built for cursive and messy hands
Structured fields
Forms, plus Queries you write yourself
Custom extractor, AI-suggested schema (Pro and up)
Tables
Strong on printed tables
Table recognition and extractors to Excel or CSV (Pro and up)
Output formats
JSON (via API or console)
Word, PDF, TXT, JSON (Excel/CSV on Pro)
Handwriting languages
English
300+ languages
Translation
Separate service (Amazon Translate)
Built in, no extra cost
API
AWS SDK with IAM
First-party token, on every plan
Pricing
Per page by feature, plus AWS usage
€19/mo for 250 pages
Privacy
Your AWS account and configuration
Encrypted, never trained on, auto-deleted
AWS Textract details and limits from its documentation and FAQs, June 2026, tested via the AWS Textract console (AnalyzeDocument: Forms, Tables and Queries). It bills per page and varies by feature, on top of your AWS usage.
Don't take our word for it
The same handwritten note, read by each tool
We wrote a realistic, messy meeting note, then ran the exact same image through AWS Textract and Handwriting OCR. No cleanup, no cherry-picking. Because we wrote it, we know exactly what it says: across the 78-word body, AWS Textract misread or dropped 28 words (64% accuracy), while Handwriting OCR read every word correctly (100%).
A messy handwritten meeting note (our own, fictional content)
Click to enlarge AWS Textract 64% word accuracy · table collapsed
Project Sym - NAes
This 14 May
Present: Priya, Marnes, Jen, Dave, Sofia R
Applogies: Tom
- Warehouse ingration shipped Vendor (Brightpath) Still waiting as API creds.
- Q3 farecast assumes 12% uplift - Marans to re-check Vs. actuals before sign- off.
- Escalation $1280. on arct #4471 restrued; process refund of
- Decision needed an Datadog renewal - seners 30 Jun.
Action items:
Owner
Taxe
Be
Due
Priya
Senl sensed Sow to Brightpath
Marcus
Referecast Q3 numbers
21 May
19 May
Jen
Close out #4471 report
16 May
Dave
Reneval vs switch memo
28 May
Next Syne Ueds 28 May, 10:30
Notes by Sofia - Sofareyes@nathgate-ops.co 0412 663 901
Handwriting OCR 100% on this note · table intact
Project Sync - Notes
Thurs 14 May
Present: Priya, Marcus, Jen, Dave, Sofia R
Apologies: Tom
- Warehouse migration slipped ~2 weeks; vendor (Brightpath) still waiting on API creds.
- Q3 forecast assumes 12% uplift - Marcus to re-check vs. actuals before sign-off.
- Escalation on acct #4471 resolved; process refund of $1280.
- Decision needed on Datadog renewal - renews 30 Jun.
Action items:
Owner | Task | Due
Priya | Send revised SOW to Brightpath | 21 May
Marcus | Reforecast Q3 numbers | 19 May
Jen | Close out #4471 refund | 16 May
Dave | Renewal vs switch memo | 28 May
Next sync: Weds 28 May, 10:30
Notes by: Sofia - sofiareyes@northgate-ops.com / 0412 663 901
misread
ambiguous / acceptable
Notes
- Same image, run through both tools, June 2026. AWS Textract misread the title, names (Marcus became 'Marnes'), and the email, dropped '~2 weeks', and moved the $1,280 out of its line.
- On the action-items table, AWS Textract's structure collapsed: the rows broke into a single column of disconnected cells (Owner / Taxe / Be / Due / Priya / ...), so owners, tasks and dates no longer line up. Handwriting OCR kept the table intact and even caught the writer's crossed-out header word.
- Handwriting OCR returned the whole page in order, with the email address and figures correct.
Source: Our own authored meeting note (fictional content, no real people). Run through both tools, June 2026.
Why the accuracy gap matters
There is a usability cliff between 64% and 100%. At roughly one word in three wrong, AWS Textract’s transcription is not a draft you tidy up, it is faster to retype, so you have either paid for OCR and still done the typing, or paid someone to. And the misreads are the costly kind: names, an email address and dollar amounts that come back plausible but wrong, pass a quick review, and land in your records as fact. For anything you feed into a spreadsheet, database or workflow, that is the difference between automation you can trust and a manual re-keying job with errors hidden inside it.
Structured data
The same note, pulled into fields
AWS Textract can extract key-values (Forms), answer Queries you write yourself, and detect Tables. We pointed all of it at the handwritten note. Handwriting OCR's custom extractor suggested the fields automatically and returned clean, production-ready JSON in under a minute: all ten fields correct (100%), including every action-item row. AWS Textract's Queries returned two of the six fields correct and could not structure the table. Field by field, here is what each returned.
| Field | AWS Textract | Handwriting OCR |
|---|---|---|
| Date | Thurs dropped, "14 May"75% confidence | Thurs 14 May |
| Attendees | Priya, Marnes, Jen, Dave, Sofia RMarcus read as "Marnes" · 54% | Priya, Marcus, Jen, Dave, Sofia R |
| Apologies | Tom100% | Tom |
| Sofareyes@nathgate-ops.comangled, missing letters · 67% | sofiareyes@northgate-ops.com | |
| Phone | 0412 663 90195% | 0412 663 901 |
| Action items (table) | One run-on blob, no rows"Owner Taxe Be Due Priya Senl sensed Sow…" · 7% | 4 structured rows: owner · task · due |
Run through AWS Textract (AnalyzeDocument: Forms and Queries) and Handwriting OCR’s custom extractor on the same image, June 2026. Confidence percentages are AWS Textract’s own. Custom extractors are on the Pro plan and up.
The action-items table is where it shows
Handwriting OCR returned four clean rows, owner, task and due date, ready to drop into a spreadsheet or database. AWS Textract’s “action items” Query could not structure it at all: it returned a single run-on blob, “Owner Taxe Be Due Priya Senl sensed Sow to Brightpath…”, at 7% confidence, with the words already misread and no rows to map. And with Queries you write each question yourself; Handwriting OCR proposed the whole schema from one sample.
Click to enlarge
Click to enlargeBoth run on the same handwritten note, June 2026. Left: AWS Textract’s Queries tab, where “action items” comes back as one blob. Right: Handwriting OCR’s custom extractor, with the action items as four separate rows.
Real switch stories
They tried AWS Textract first
"I tested the custom extractor on a large document, and the accuracy is surprisingly high. Much better than AWS Textract."
Jakub M.
Senior Product Manager, ERP
"In researching OCR for transcribing handwritten student essays, I found your accuracy outperforms AWS Textract and Azure Document Intelligence."
Reena C.
Founder, AI EdTech platform
"I had six PDF-scanned handwritten journals in Dutch to transcribe. Handwriting OCR was the best I've found. I tried AWS Textract too, but it was poor."
Simon
Digitising handwritten Dutch journals
Accuracy
Reads the handwriting AWS Textract trips on
AWS Textract is tuned for printed text and does a fair job on neat English handwriting. Connected cursive, faded ink and older hands are where it slips. Handwriting OCR is built specifically for connected and historical handwriting, so the names, figures and details come back right the first time.
- ✓ Cursive, mixed print-and-cursive, and hands from the 1940s onward
- ✓ Keeps punctuation, capitalisation and layout instead of guessing
- ✓ Handles several writers in one document without reconfiguring
Structured data
Handwritten forms, straight to a spreadsheet
With AWS Textract you choose between Forms, Queries you author by hand, and Tables, then wire the JSON together yourself. Handwriting OCR's custom extractors suggest the fields from a sample and pull named values and nested tables into proper columns, exporting to Excel or CSV, so 200 handwritten forms become 200 tidy rows. (Available on the Pro plan and up.)
- ✓ Let the AI suggest the schema, then run thousands of forms through it
- ✓ Nested tables come back as real rows, not one merged blob
- ✓ Export to Excel, CSV or JSON for your own systems
No cloud plumbing
A first-party API on every plan
AWS Textract means an AWS account, IAM roles, S3 buckets and the SDK before your first call. Handwriting OCR’s API is our own and simple: a Bearer token, webhooks, and structured JSON or direct Excel and CSV. It’s on every plan, including the free trial, so you can automate from day one.
Private by default
Your documents stay yours
Files are encrypted in transit and at rest, never used to train our models, and auto-deleted on a schedule you set (default 7 days, anywhere from 15 minutes to 14 days). Sensitive notes, records and forms never become someone’s training data.
Read it in any language
Built-in translation, no extra cost
Transcribe a note in French, German, Spanish or any of 300+ languages, then translate it to English (or back) in the same workflow. It’s included on every plan, with no second service to wire in and no extra charge.
Being fair
When AWS Textract is the better pick
AWS Textract is a serious, capable service, and for some projects it is genuinely the right choice. If one of these is you, it earns its place, and that's an honest answer, not a sales one.
- ✓Your documents are mostly printed or typed forms, invoices and IDs
- ✓You are already deep in AWS and want native S3, Lambda and IAM integration
- ✓You need the specialised models (AnalyzeExpense, AnalyzeID, Lending) for US financial and ID documents
- ✓You want raw per-page cloud pricing at very large scale
Pricing
Plans for every project
Pay-as-you-go credits or monthly subscriptions. Cancel any time.
Starter
250 pages / month
£19 $19 €19 £16 $16 €16 /month
Billed monthlyBilled annually
- AI-enhanced formatting
- Export to Markdown (plain text)
- Export to Microsoft Word
- Two-factor authentication
- API access
- Renews monthly, cancel any time
- Additional pages: £6.00 $8.00 €6.50 £5.50 $6.00 €6.00 / 100 pages
Pro
1,000 pages / month
£49 $59 €59 £41 $50 €50 /month
Billed monthlyBilled annually
- Everything in Starter, plus:
- Export tables to Microsoft Excel
- Custom extractors
- Additional pages: £5.00 $6.00 €5.00 £5.00 $5.00 €5.00 / 100 pages
Business
10,000 pages / month
£399 $499 €490 £333 $416 €409 /month
Billed monthlyBilled annually
- Everything in Pro, plus:
- Up to 5 team members
- Configurable audit logging
- Additional pages: £4.00 $5.00 €4.50 £3.50 $4.00 €4.00 / 100 pages
For higher volumes, options for offline deployment, or any other custom requirements, please contact us.
FAQ
AWS Textract alternative: common questions
Anything else? Get in touch and we'll answer right away.
What is Handwriting OCR?
Handwriting OCR is an AI service that turns photos and scans of handwritten documents into clean, editable text and structured data. It reads cursive, messy and historical hands, and pulls handwritten forms, notes and tables into named fields, exporting to Word, PDF, plain text and JSON, plus Excel and CSV on Pro and up.
Who is behind Handwriting OCR?
Handwriting OCR is built by a small independent UK team. The company was founded in London in 2023, runs as a focused product, and never trains its models on your documents.
How is Handwriting OCR different from AWS Textract?
AWS Textract is a cloud service built for printed forms, tables and invoices at scale, wired into S3, IAM and the AWS SDK. Handwriting OCR is built for handwriting: it reads cursive and messy hands, suggests an extraction schema for you, returns structured JSON or Excel, and has a simple first-party API on every plan with no AWS setup.
Does AWS Textract read handwriting?
Yes, AWS Textract detects English handwriting as well as printed text, and it is strong on printed forms and tables. On real cursive and messy notes it slips: in our June 2026 test it misread names, mangled an email, dropped words and could not structure a handwritten action-items table. The comparison higher up this page shows the same note run through both tools.
Can I extract handwritten forms to structured fields or Excel?
Yes. Handwriting OCR custom extractors pull named fields (and nested tables) out of handwritten forms and return them as JSON, Excel or CSV. You can let the AI suggest the fields from a sample, so a working extractor is ready in about a minute, with no query strings to write by hand. Custom extractors and table extraction are on the Pro plan and up.
Is my handwriting kept private?
Your documents are encrypted in transit and at rest, are never used to train our models, and are auto-deleted on a schedule you control (default 7 days, configurable from 15 minutes to 14 days). We do not currently offer HIPAA BAAs; for compliance questions, get in touch.
Try it on your own documents
Bring the handwriting AWS Textract couldn’t read.
Free trial credits, no credit card, no AWS account. Upload the page you’re stuck on and see the text and fields come back in seconds.
Our experience
What people actually switch from AWS Textract for
Most people do not arrive looking for an “AWS Textract alternative” in the abstract. They arrive with a specific job AWS Textract got wrong: a stack of handwritten intake forms, field notes, survey cards or meeting minutes that came back as scrambled text and a table that fell apart.
AWS Textract is genuinely good at what it was built for. On printed invoices, typed forms and structured tables, at AWS scale, it is a strong service, and its specialised models for expenses and IDs are useful. The wall is handwriting. Detecting that a page contains handwriting is not the same as reading cursive accurately, and once the values are misread, every downstream Form, Query and Table inherits the error. That is exactly what the comparison above shows: a believable handwritten note where the names, the email and the action-items table all came back wrong, with AWS Textract’s own confidence scores as low as 7%.
Handwriting OCR is built around that gap. It reads cursive and messy hands, keeps the layout, and turns a handwritten form into named fields and real table rows, with a schema it can suggest for you from a single sample rather than queries you write by hand. It never trains on your files, it runs from a first-party API on every plan with no S3 or IAM to configure, and it is made by a small UK team that treats handwriting recognition as the whole product.
If you have a tricky page or a large batch, the fastest way to judge it is to try it free on the exact document AWS Textract struggled with.