What matters
- Vietnamese-specific requirements are the real axis, not price. The comparison actually rests on three things: accuracy on Vietnamese text (especially handwriting), on-premise deployment, and whether data stays in-country, headline per-page API cost is a distraction.
- GreenNode IDP is the only option built around all three. 99% accuracy on printed Vietnamese, 94% on handwritten, on-premise support, and classification + extraction + crosscheck bundled into one workflow instead of stitched-together APIs.
- AWS, Azure, and Google all undercut on raw OCR (~$1.50/1K pages) but the real cost is structured extraction, which runs 20–40x higher and none of them classify documents or reconcile across files without custom engineering work.
- Vietnam's 2025 Personal Data Protection Law makes data residency non-negotiable for regulated sectors, and none of the three global platforms can satisfy it without a bolted-on compliance layer built in-house.
- The deployment-speed gap is the sharpest differentiator: GreenNode claims 2–3 days via no-code templates versus "several weeks" for the global platforms' custom model training, and the 40–80 hours of hidden engineering work is often the largest real cost once volume scales up.
Amazon Textract, Azure Document Intelligence, Google Document AI, and GreenNode IDP all solve the same problem: extracting structured data from documents. But for Vietnamese businesses, the choice isn't just about price or API call volume. Three factors usually decide it: accuracy with Vietnamese text, on-premise deployment capability, and whether data can stay within the country to meet compliance requirements.
Among this group, the three international platforms — Textract, Azure Document Intelligence, and Google Document AI — are all strong at document automation, but none is specifically optimized for Vietnamese and none supports on-premise deployment in Vietnam. So when evaluating them for a real use case, pricing is only part of the story. What matters more is which platform fits your data, language, and operational requirements.
GreenNode IDP: Overview, Pricing, and Use Cases
Overview of GreenNode IDP
GreenNode IDP is Vietnam's intelligent document processing platform, combining traditional OCR with a large language model (LLM) and a large vision-language model (VLM). Unlike the other three platforms, which are collections of separate APIs businesses must stitch together themselves, GreenNode IDP packages all four layers into a single processing flow: document classification, field extraction, crosscheck & risk scoring, and output normalization into each customer's own schema (JSON/XML/flat). The platform includes models fine-tuned for specific document types (national ID cards, passports, land-use certificates, vehicle registration, banking documents, insurance documents), plus the ability to define new form templates through a drag-and-drop interface without waiting for the vendor to retrain a model.
On performance, GreenNode IDP processes documents in under 3 seconds per page on RTX 4090 GPU infrastructure and under 1 second per page on NVIDIA H100 GPUs, with a batch & parallel architecture that keeps throughput stable even when document volume spikes. On security, data is encrypted in transit, operating environments are fully isolated between customers (multi-tenant isolation), and the platform commits to never using or storing customer data beyond the scope of authorized processing.
GreenNode IDP Pricing
GreenNode IDP doesn't charge per API call like the other three platforms. Pricing is quoted based on volume and deployment model (cloud or on-premise), typically preceded by a free demo and pilot phase before committing to a larger budget. This approach helps businesses avoid the risk of bills spiking when structured-extraction features are turned on — something common with model-based API pricing, where the gap between raw OCR and structured extraction can reach 20-40x.
What GreenNode IDP Is Best For
GreenNode IDP is built for Vietnamese businesses processing large volumes of documents containing Vietnamese handwriting (prescriptions, incident reports, handwritten records), that need data stored locally or deployed on-premise for compliance reasons, and that want business teams to independently onboard new form templates without depending on an AI engineering team every time a form changes.
Read more: Automating Health Insurance Claims: From 30 Minutes to Under 5 Minutes per Case with IDP
AWS Textract: Overview, Pricing, and Use Cases
Overview of AWS Textract
Amazon Textract is AWS's OCR/IDP service, structured around several separate APIs: DetectDocumentText (raw OCR), AnalyzeDocument (individually enabled Forms/Tables/Queries/Signatures features), AnalyzeExpense (invoices/receipts), AnalyzeID (identity documents), and AnalyzeLending (loan documents). It's a popular choice for businesses already running on AWS infrastructure, since it integrates directly with S3, Lambda, and other AWS services without needing a middle layer.
Worth noting: Textract doesn't automatically classify or crosscheck data across documents within the same case file — this is business logic you have to build on top of the API's output yourself. For case files with multiple document types (such as a contract bundled with ID documents and invoices), you need to chain multiple API calls and build your own reconciliation and completeness-checking steps.
AWS Textract Pricing
- DetectDocumentText (raw OCR): $1.50 per 1,000 pages.
- AnalyzeDocument with Forms: approximately $50 per 1,000 pages.
- Tables: approximately $15 per 1,000 pages (additive if enabled alongside Forms).
- AnalyzeExpense: approximately $10 per 1,000 pages.
The gap between raw OCR and full structured extraction can reach more than 40x depending on which features are combined — and this is pure API cost, not counting engineering time to build and maintain the pipeline.
What AWS Textract Is Best For
Businesses that already have infrastructure and an engineering team running on AWS, that process English-language documents in standardized formats (US-style invoices and receipts), and that are prepared to build their own ingestion pipeline, retry logic, and API orchestration. Textract has no model specifically tuned for Vietnamese handwriting, and data is processed on infrastructure located outside Vietnam — a point worth weighing for case files containing sensitive personal data under current regulations.
Azure Document Intelligence: Overview, Pricing, and Use Cases
Overview of Azure Document Intelligence
Azure AI Document Intelligence (renamed from Form Recognizer in 2023) is Microsoft's document AI service, split into tiers: Read (OCR), Layout, Prebuilt models (invoice, receipt, ID, W-2, health insurance card, etc.), Custom Classification, and Custom Extraction. Its strength is native integration with the Microsoft ecosystem — Power Automate, Logic Apps, and SharePoint — making it a fit for businesses already running internal workflows on Microsoft 365.
For form types not covered by a Prebuilt model, businesses need to train their own Custom Extraction model — a process that adds training costs (around $3/hour after the first 10 free hours each month) and time to prepare sample data, unlike GreenNode IDP's instant template-onboarding mechanism.
Azure Document Intelligence Pricing
- Read (raw OCR): $1.50 per 1,000 pages.
- Layout and Prebuilt models: $10 per 1,000 pages.
- Custom Classification: $3 per 1,000 pages.
- Custom Extraction: $30 per 1,000 pages.
Azure also offers a discounted commitment tier for very high volumes (millions of pages per month), but this isn't shown publicly on the default pricing page and requires direct contact to negotiate.
What Azure Document Intelligence Is Best For
Businesses already invested in the Microsoft 365/Azure ecosystem that need to plug document AI into existing automation workflows (Power Automate, SharePoint). Like Textract, Azure Document Intelligence has no model optimized specifically for Vietnamese handwriting, and custom extraction for specialized forms requires its own training process, taking more time than GreenNode IDP's template-onboarding mechanism.
Google Document AI: Overview, Pricing, and Use Cases
Overview of Google Document AI
Google Document AI is part of Google Cloud, offering processors including Enterprise Document OCR, Layout Parser, Form Parser, and Custom Extractor. The platform is strong at multilingual processing thanks to Google's underlying multilingual models, and offers clear volume-based discounts — a fit for businesses processing documents from multiple markets at once.
Like Azure and AWS, onboarding a new form template with Custom Extractor requires labeling training data and deploying your own processor version (each deployed version billed hourly) — not the instant draw-label-publish workflow GreenNode IDP offers.
Google Document AI Pricing
- Enterprise Document OCR: $1.50 per 1,000 pages.
- Layout Parser: $10 per 1,000 pages.
- Form Parser and Custom Extractor: $30 per 1,000 pages for 1-1,000,000 pages/month, dropping to $20 per 1,000 pages above the 1 million pages/month mark.
What Google Document AI Is Best For
Businesses already on Google Cloud, processing multilingual documents at very large scale and looking to take advantage of volume discounts. Like the other two international platforms, Google Document AI doesn't yet have a model specifically tuned for Vietnamese handwriting and doesn't support on-premise deployment in Vietnam.
The Hidden Costs of Self-Building on International Platforms
API pricing is only part of the total cost. With all three international platforms, businesses need to add: engineering time to build and maintain the pipeline (typically 40-80 hours for the initial rollout, based on common industry estimates), supporting infrastructure costs (temporary storage, processing queues, retry mechanisms), and data-labeling costs every time a custom model needs training for a new form.
This is the part most often left out of pure "price per page" comparisons — yet it's typically the largest real operating cost at a scale of hundreds of thousands to millions of pages per month.
Quick Comparison of Document Processing Solutions
| Criteria | GreenNode IDP | AWS Textract | Azure Document Intelligence | Google Document AI |
|---|---|---|---|---|
| Raw OCR price | Quoted by volume | $1.50 per 1,000 pages | $1.50 per 1,000 pages | $1.50 per 1,000 pages |
| Structured extraction price | All-inclusive, not billed per API | ~$15-50 per 1,000 pages (Tables/Forms) | $10-30 per 1,000 pages (Prebuilt/Custom) | $30 per 1,000 pages (drops to $20 above 1 million pages/month) |
| Printed-text accuracy (field level) | 99% | Not published specifically for Vietnamese | Not published specifically for Vietnamese | Not published specifically for Vietnamese |
| Vietnamese handwriting accuracy | 94% | No model specialized for Vietnamese | No model specialized for Vietnamese | No model specialized for Vietnamese |
| Processing speed | Under 2.5 sec/page (RTX 4090), under 1 sec/page (H100) | Depends on customer-built AWS infrastructure | Depends on customer-built Azure infrastructure | Depends on customer-built GCP infrastructure |
| Built-in classification & crosscheck | Yes | No, must be built separately | No, must be built separately | No, must be built separately |
| On-premise deployment | Yes | No (AWS cloud only) | No (Azure cloud only) | No (GCP cloud only) |
| Self-service new template onboarding, no code | Yes, under 30 minutes | Requires training a separate custom model | Requires training a separate custom model | Requires training a separate custom model |
| Data stored in Vietnam | Yes | No | No | No |
| Time to first experience | 2-3 days | Depends on internal pipeline build time (typically several weeks) | Depends on internal pipeline build time (typically several weeks) | Depends on internal pipeline build time (typically several weeks) |
Which Platform Is Right for You?
- Processing a lot of Vietnamese handwriting, need data to stay local or on-premise, and want your business team to be self-sufficient when forms change → GreenNode IDP.
- Already running on AWS, documents are mostly standardized English-language forms, and you have your own engineering team to build the pipeline → AWS Textract.
- Already invested in the Microsoft 365/Power Platform ecosystem and need to plug into existing automation workflows → Azure Document Intelligence.
- Already on Google Cloud, processing multilingual documents at very large scale, and prioritizing volume-based discounts → Google Document AI.
For most Vietnamese businesses processing handwritten documents and subject to data residency requirements under the 2025 Personal Data Protection Law, all three international platforms require an additional layer of customization, a self-built pipeline, and separate infrastructure to meet those needs — while this is a capability GreenNode IDP has built in from the start.
If you're weighing these options for your own document volume, the GreenNode team is ready to run a live demo on your actual data to compare accuracy and processing time. Contact our advisory team, or send details about your business's use case to support@greennode.ai or hotline 1900 1549.
