Preparing for first start
Setting up storage and offline tools. This can take a bit longer after installing the app or clearing app data.
One Platform. Complete Control.
Powered byTotal Access Platform
Preparing for first start
Setting up storage and offline tools. This can take a bit longer after installing the app or clearing app data.
One Platform. Complete Control.
Preparing for first start
Setting up storage and offline tools. This can take a bit longer after installing the app or clearing app data.
One Platform. Complete Control.
The AI-powered OCR API that reads invoices, receipts, bank statements, IDs, and 40+ document types — with confidence scores, custom models, and multi-language support.
Total Vision is a developer-first document AI platform. Send a PDF or photo, get back structured, validated JSON. Extract line items, classify documents, split multi-page batches, and train custom models on your own document layouts — all via a single REST API.
Generic OCR gives you text. Your business needs structured data — vendor names, line items, tax amounts, GL codes. Building that pipeline yourself takes months and never quite handles every layout.
Writing regex and template rules for every supplier's invoice format. Then a supplier changes their layout and you start over.
Generic OCR handles the easy 20%. The remaining 80% — rotated scans, handwriting, multi-page tables, mixed languages — is where your team gets stuck.
A dedicated ML engineer, annotation tooling, GPU inference, and ongoing maintenance. Or: one API call at R1.75 per page.
Total Vision handles the entire document processing pipeline — from raw image to validated, structured JSON — so your team can focus on your core product, not OCR edge cases.
Get 30 free creditsOne API call handles the entire pipeline — no chaining multiple services together.
Send a PDF, image, or base64-encoded file via POST /api/v1/vision/extract. Multi-page documents supported.
AI identifies the document type (invoice, receipt, bank statement, ID, etc.) and routes to the right extraction model.
Field-level extraction with confidence scores and bounding boxes. Line items, tables, handwriting, and multi-language text.
Cross-check against your master data, business rules, and custom validation logic. Returns structured JSON ready for your system.
The four pillars of document AI — extraction, classification, splitting, and cropping — in a single API. Plus custom models, chaining, and continuous learning.
Extract fields, line items, tables, and key-value pairs from any document format. Per-field confidence scores and bounding boxes for verifiable accuracy.
Automatically identify document type — invoice, receipt, bank statement, ID, contract — and route to the correct extraction model. No manual sorting.
Detect document boundaries in multi-page uploads and split into individual records. Perfect for batch scanning and email attachments with multiple documents.
Isolate multiple documents scanned on a single page. Each item is cropped into a standalone file and processed individually.
Chain pre-processing and extraction models in a single API call. Classify → split → crop → extract, all in one request. No additional cost.
Train extraction models on your own document layouts. Upload sample documents, define your fields, and the AI learns your format — no ML expertise required.
The AI improves with every correction. Upload corrected extractions to build a knowledge base of edge cases, turning rare exceptions into automated successes.
Read handwritten notes, signatures, and annotations alongside printed text. Supports cursive and print styles across multiple scripts.
From invoices to ID documents to customs declarations — our models are pre-trained on millions of South African and international documents. No training required for standard types.
Merchant, line items, totals, tax, payment method
Vendor, bill-to, line items, totals, due date, payment terms
Account, opening/closing balance, transactions, deposits, withdrawals
PO number, supplier, line items, delivery date, payment terms
Goods received, quantities, supplier, PO reference
Supplier, amounts, tax, GL suggestions, due date
Reference, original invoice, amounts, tax adjustments
Statement period, opening balance, transactions, closing balance
ID number, full name, date of birth, nationality, expiry
MRZ, passport number, name, nationality, issue/expiry dates
License number, classes, address, expiry, restrictions
Utility bills, council rates, bank letters — address & date extraction
Contact, skills, work history, education, certifications
Parties, start/end dates, terms, financials, clauses
Applicant, loan amount, type, income, credit score, status
Claim ID, policy number, amount, incident details, adjuster notes
HS codes, origin, destination, value, commodity descriptions
Shipper, consignee, port of loading/discharge, container details
Patient, diagnosis codes, medications, physician notes (POPIA compliant)
Shipment ID, origin, destination, carrier, tracking, items
Consignor, consignee, goods, quantity, freight charges
Order reference, items packed, quantities, destination
Product, quantity, reason, warehouse, reference
Original order, reason, item, condition, refund amount
Quotation reference, items, terms, validity
Customer, items, quantities, delivery, pricing
Job reference, operations, labour, materials, status
Requestor, items, quantities, budget, approval
Client, period, line items, rates, totals
VAT201, EMP201, IRP5, EMP501 — figures, periods, totals
Employee, period, earnings, deductions, net pay, YTD
Define your own fields — our AI learns your layout
Don't see your document type? Define custom fields and our AI learns your layout.
Request a custom modelEnterprise document processing isn't just about reading text. It's about ensuring only valid, enriched data enters your downstream systems.
Parse complex multi-row tables with column headers, merged cells, and subtotals. Each line item returned as a structured object with its own confidence score.
Process documents in any language — including RTL scripts (Arabic, Hebrew), CJK (Chinese, Japanese, Korean), and African languages. Automatic language detection.
Every extracted field includes a confidence score (0-1) and pixel-level bounding box coordinates. Build human-in-the-loop review for low-confidence fields only.
Validate extracted vendor names against your supplier master, GL codes against your chart of accounts, and tax rates against SARS tables. Reject invalid data before it enters your ERP.
Upload corrected documents to build a retrieval-augmented knowledge base. The AI references past corrections to handle non-standard layouts — turning rare exceptions into automated successes.
Detect tampered invoices, forged receipts, and manipulated bank statements. Checks include metadata analysis, pixel-level forensics, and cross-referencing against known document templates.
Track your STP rate — the percentage of documents processed with no human intervention. Dashboard shows accuracy trends, exception types, and processing time per document type.
Every processed document is archived with its extracted data. Search across both extracted and non-extracted text. Full audit trail with timestamps and user actions for compliance.
Get notified when extraction completes. Configure webhooks per document type, per confidence threshold. HMAC-signed payloads with automatic retries and dead-letter queue.
No SDKs to install, no models to train, no infrastructure to manage. Send a document, get JSON. It's that simple.
Total Vision integrates with the systems you already use — or use it standalone via API. Export to any accounting platform, ERP, or downstream system.
Xero, Sage, QuickBooks, Pastel, Draftworx — export extracted data directly into your accounting software with source documents attached.
SAP, Oracle NetSuite, Microsoft Dynamics, Workday, Coupa — push validated, enriched data into your ERP via API or webhook.
Stitch API integration for live bank feeds. OFX, QIF, MT940, CSV parsers for FNB, Standard Bank, ABSA, Nedbank, Capitec, Investec.
Get notified when extraction completes. HMAC-signed payloads, configurable per document type and confidence threshold. Automatic retries with DLQ.
Node.js, Python, PHP SDKs. OpenAPI/Swagger spec. Postman collection. Interactive API explorer in the developer portal.
Expose document extraction to AI agents via MCP adapter. Claude, Cursor, and custom AI agents can call the extraction API as a tool.
More document types, SA-localised, credit-based pricing with no expiry, and built-in integration with the full Total Access platform.
| Feature | Total Vision TotalAccess | Mindee | Rossum | DocuPipe | ClerkIQ |
|---|---|---|---|---|---|
| Document Processing | |||||
| Pre-trained document types | 40+ | 15+ | 20+ | 20+ | 5 (bank statements) |
| Custom model training | |||||
| Document classification | |||||
| Document splitting | |||||
| Auto-crop multi-doc pages | |||||
| Model chaining (1 API call) | |||||
| Handwriting recognition | |||||
| Table / line-item extraction | |||||
| Confidence scores & bounding boxes | |||||
| RAG for edge cases | |||||
| Document fraud detection | |||||
| Language & Localization | |||||
| Languages supported | 276 | Any | 276 | Any | 1 (English) |
| RTL script support (Arabic, Hebrew) | |||||
| SA bank statement specialization | |||||
| SARS VAT / tax form extraction | |||||
| SA ID document extraction | |||||
| African language support | |||||
| Validation & Automation | |||||
| Cross-validation with master data | |||||
| GL code suggestion | |||||
| Tax code computation | |||||
| PO → invoice matching | |||||
| Straight-through processing metrics | |||||
| Approval workflow automation | |||||
| Vendor email automation | |||||
| Integration & Developer Experience | |||||
| REST API | |||||
| Webhook callbacks | |||||
| SDKs (Node, Python, PHP) | |||||
| OAuth 2.0 | |||||
| Sandbox environment | |||||
| OpenAPI / Swagger spec | |||||
| AI Agent Bridge (MCP) | |||||
| Accounting platform integrations | 5+ | 0 | 6+ | 0 | 0 |
| Pricing & Deployment | |||||
| Credit-based pricing | |||||
| Credits never expire | N/A | ||||
| Free trial credits | 30 | 50 | 14 days | 20 | 30 |
| SA-hosted data | |||||
| POPIA compliant | |||||
| Enterprise self-hosted option | |||||
Your documents contain sensitive financial and personal data. We treat them accordingly.
All document processing is POPIA-compliant. Data is hosted in South Africa, with documented processing records and data subject access support.
TLS 1.3 in transit, AES-256 at rest. Documents are encrypted the moment they reach our servers and decrypted only during processing.
Your documents never leave South African soil. All processing and storage is in SA-based data centres with local redundancy.
Authenticate with scoped API keys or OAuth 2.0 apps. Per-key rate limits, per-scope permissions, and full audit logging of every API call.
Every extraction, correction, and API call is logged with timestamp, user, IP, and document hash. Exportable for compliance audits.
Configure retention per document type — from immediate deletion after extraction to 10-year archival for audit requirements. You control your data lifecycle.
Every uploaded file is scanned for malware before processing. Files that fail scanning are quarantined and the API returns a safe error response.
Detect tampered documents with metadata analysis, pixel-level forensics, and template cross-referencing. Flag suspicious documents for manual review.
“We replaced our in-house OCR pipeline with Total Vision and cut our processing time by 90%. The SA bank statement extraction alone saved us two full-time data capturers.”
“The confidence scores are a game-changer. We only review fields below 85% confidence, which means 70% of our invoices go straight through with no human touch.”
“We integrated the API in an afternoon. The Node.js SDK and the interactive API explorer made it trivial. Our supplier statements are now processed before our morning coffee.”
30 free credits on signup. No credit card required. Credits never expire.
The AI-powered OCR API that reads invoices, receipts, bank statements, IDs, and 40+ document types — with confidence scores, custom models, and multi-language support.
Total Vision is a developer-first document AI platform. Send a PDF or photo, get back structured, validated JSON. Extract line items, classify documents, split multi-page batches, and train custom models on your own document layouts — all via a single REST API.
Generic OCR gives you text. Your business needs structured data — vendor names, line items, tax amounts, GL codes. Building that pipeline yourself takes months and never quite handles every layout.
Writing regex and template rules for every supplier's invoice format. Then a supplier changes their layout and you start over.
Generic OCR handles the easy 20%. The remaining 80% — rotated scans, handwriting, multi-page tables, mixed languages — is where your team gets stuck.
A dedicated ML engineer, annotation tooling, GPU inference, and ongoing maintenance. Or: one API call at R1.75 per page.
Total Vision handles the entire document processing pipeline — from raw image to validated, structured JSON — so your team can focus on your core product, not OCR edge cases.
Get 30 free creditsOne API call handles the entire pipeline — no chaining multiple services together.
Send a PDF, image, or base64-encoded file via POST /api/v1/vision/extract. Multi-page documents supported.
AI identifies the document type (invoice, receipt, bank statement, ID, etc.) and routes to the right extraction model.
Field-level extraction with confidence scores and bounding boxes. Line items, tables, handwriting, and multi-language text.
Cross-check against your master data, business rules, and custom validation logic. Returns structured JSON ready for your system.
The four pillars of document AI — extraction, classification, splitting, and cropping — in a single API. Plus custom models, chaining, and continuous learning.
Extract fields, line items, tables, and key-value pairs from any document format. Per-field confidence scores and bounding boxes for verifiable accuracy.
Automatically identify document type — invoice, receipt, bank statement, ID, contract — and route to the correct extraction model. No manual sorting.
Detect document boundaries in multi-page uploads and split into individual records. Perfect for batch scanning and email attachments with multiple documents.
Isolate multiple documents scanned on a single page. Each item is cropped into a standalone file and processed individually.
Chain pre-processing and extraction models in a single API call. Classify → split → crop → extract, all in one request. No additional cost.
Train extraction models on your own document layouts. Upload sample documents, define your fields, and the AI learns your format — no ML expertise required.
The AI improves with every correction. Upload corrected extractions to build a knowledge base of edge cases, turning rare exceptions into automated successes.
Read handwritten notes, signatures, and annotations alongside printed text. Supports cursive and print styles across multiple scripts.
From invoices to ID documents to customs declarations — our models are pre-trained on millions of South African and international documents. No training required for standard types.
Merchant, line items, totals, tax, payment method
Vendor, bill-to, line items, totals, due date, payment terms
Account, opening/closing balance, transactions, deposits, withdrawals
PO number, supplier, line items, delivery date, payment terms
Goods received, quantities, supplier, PO reference
Supplier, amounts, tax, GL suggestions, due date
Reference, original invoice, amounts, tax adjustments
Statement period, opening balance, transactions, closing balance
ID number, full name, date of birth, nationality, expiry
MRZ, passport number, name, nationality, issue/expiry dates
License number, classes, address, expiry, restrictions
Utility bills, council rates, bank letters — address & date extraction
Contact, skills, work history, education, certifications
Parties, start/end dates, terms, financials, clauses
Applicant, loan amount, type, income, credit score, status
Claim ID, policy number, amount, incident details, adjuster notes
HS codes, origin, destination, value, commodity descriptions
Shipper, consignee, port of loading/discharge, container details
Patient, diagnosis codes, medications, physician notes (POPIA compliant)
Shipment ID, origin, destination, carrier, tracking, items
Consignor, consignee, goods, quantity, freight charges
Order reference, items packed, quantities, destination
Product, quantity, reason, warehouse, reference
Original order, reason, item, condition, refund amount
Quotation reference, items, terms, validity
Customer, items, quantities, delivery, pricing
Job reference, operations, labour, materials, status
Requestor, items, quantities, budget, approval
Client, period, line items, rates, totals
VAT201, EMP201, IRP5, EMP501 — figures, periods, totals
Employee, period, earnings, deductions, net pay, YTD
Define your own fields — our AI learns your layout
Don't see your document type? Define custom fields and our AI learns your layout.
Request a custom modelEnterprise document processing isn't just about reading text. It's about ensuring only valid, enriched data enters your downstream systems.
Parse complex multi-row tables with column headers, merged cells, and subtotals. Each line item returned as a structured object with its own confidence score.
Process documents in any language — including RTL scripts (Arabic, Hebrew), CJK (Chinese, Japanese, Korean), and African languages. Automatic language detection.
Every extracted field includes a confidence score (0-1) and pixel-level bounding box coordinates. Build human-in-the-loop review for low-confidence fields only.
Validate extracted vendor names against your supplier master, GL codes against your chart of accounts, and tax rates against SARS tables. Reject invalid data before it enters your ERP.
Upload corrected documents to build a retrieval-augmented knowledge base. The AI references past corrections to handle non-standard layouts — turning rare exceptions into automated successes.
Detect tampered invoices, forged receipts, and manipulated bank statements. Checks include metadata analysis, pixel-level forensics, and cross-referencing against known document templates.
Track your STP rate — the percentage of documents processed with no human intervention. Dashboard shows accuracy trends, exception types, and processing time per document type.
Every processed document is archived with its extracted data. Search across both extracted and non-extracted text. Full audit trail with timestamps and user actions for compliance.
Get notified when extraction completes. Configure webhooks per document type, per confidence threshold. HMAC-signed payloads with automatic retries and dead-letter queue.
No SDKs to install, no models to train, no infrastructure to manage. Send a document, get JSON. It's that simple.
Total Vision integrates with the systems you already use — or use it standalone via API. Export to any accounting platform, ERP, or downstream system.
Xero, Sage, QuickBooks, Pastel, Draftworx — export extracted data directly into your accounting software with source documents attached.
SAP, Oracle NetSuite, Microsoft Dynamics, Workday, Coupa — push validated, enriched data into your ERP via API or webhook.
Stitch API integration for live bank feeds. OFX, QIF, MT940, CSV parsers for FNB, Standard Bank, ABSA, Nedbank, Capitec, Investec.
Get notified when extraction completes. HMAC-signed payloads, configurable per document type and confidence threshold. Automatic retries with DLQ.
Node.js, Python, PHP SDKs. OpenAPI/Swagger spec. Postman collection. Interactive API explorer in the developer portal.
Expose document extraction to AI agents via MCP adapter. Claude, Cursor, and custom AI agents can call the extraction API as a tool.
More document types, SA-localised, credit-based pricing with no expiry, and built-in integration with the full Total Access platform.
| Feature | Total Vision TotalAccess | Mindee | Rossum | DocuPipe | ClerkIQ |
|---|---|---|---|---|---|
| Document Processing | |||||
| Pre-trained document types | 40+ | 15+ | 20+ | 20+ | 5 (bank statements) |
| Custom model training | |||||
| Document classification | |||||
| Document splitting | |||||
| Auto-crop multi-doc pages | |||||
| Model chaining (1 API call) | |||||
| Handwriting recognition | |||||
| Table / line-item extraction | |||||
| Confidence scores & bounding boxes | |||||
| RAG for edge cases | |||||
| Document fraud detection | |||||
| Language & Localization | |||||
| Languages supported | 276 | Any | 276 | Any | 1 (English) |
| RTL script support (Arabic, Hebrew) | |||||
| SA bank statement specialization | |||||
| SARS VAT / tax form extraction | |||||
| SA ID document extraction | |||||
| African language support | |||||
| Validation & Automation | |||||
| Cross-validation with master data | |||||
| GL code suggestion | |||||
| Tax code computation | |||||
| PO → invoice matching | |||||
| Straight-through processing metrics | |||||
| Approval workflow automation | |||||
| Vendor email automation | |||||
| Integration & Developer Experience | |||||
| REST API | |||||
| Webhook callbacks | |||||
| SDKs (Node, Python, PHP) | |||||
| OAuth 2.0 | |||||
| Sandbox environment | |||||
| OpenAPI / Swagger spec | |||||
| AI Agent Bridge (MCP) | |||||
| Accounting platform integrations | 5+ | 0 | 6+ | 0 | 0 |
| Pricing & Deployment | |||||
| Credit-based pricing | |||||
| Credits never expire | N/A | ||||
| Free trial credits | 30 | 50 | 14 days | 20 | 30 |
| SA-hosted data | |||||
| POPIA compliant | |||||
| Enterprise self-hosted option | |||||
Your documents contain sensitive financial and personal data. We treat them accordingly.
All document processing is POPIA-compliant. Data is hosted in South Africa, with documented processing records and data subject access support.
TLS 1.3 in transit, AES-256 at rest. Documents are encrypted the moment they reach our servers and decrypted only during processing.
Your documents never leave South African soil. All processing and storage is in SA-based data centres with local redundancy.
Authenticate with scoped API keys or OAuth 2.0 apps. Per-key rate limits, per-scope permissions, and full audit logging of every API call.
Every extraction, correction, and API call is logged with timestamp, user, IP, and document hash. Exportable for compliance audits.
Configure retention per document type — from immediate deletion after extraction to 10-year archival for audit requirements. You control your data lifecycle.
Every uploaded file is scanned for malware before processing. Files that fail scanning are quarantined and the API returns a safe error response.
Detect tampered documents with metadata analysis, pixel-level forensics, and template cross-referencing. Flag suspicious documents for manual review.
“We replaced our in-house OCR pipeline with Total Vision and cut our processing time by 90%. The SA bank statement extraction alone saved us two full-time data capturers.”
“The confidence scores are a game-changer. We only review fields below 85% confidence, which means 70% of our invoices go straight through with no human touch.”
“We integrated the API in an afternoon. The Node.js SDK and the interactive API explorer made it trivial. Our supplier statements are now processed before our morning coffee.”
30 free credits on signup. No credit card required. Credits never expire.