Mindee OCR API vs IronOCR: .NET OCR Library

Mindee Invoice Parsing Pattern

The Mindee invoice extraction flow is straightforward when cloud transmission is acceptable:

using Mindee;
using Mindee.Input;
using Mindee.Product.Invoice;

public class MindeeInvoiceService
{
    private readonly MindeeClient _client;

    public MindeeInvoiceService(string apiKey)
    {
        // API key authenticates with Mindee cloud
        _client = new MindeeClient(apiKey);
    }

    public async Task<InvoiceData> ParseInvoiceAsync(string filePath)
    {
        // Document leaves your infrastructure at this line
        var inputSource = new LocalInputSource(filePath);

        // Full document transmitted to Mindee — bank details, tax IDs, line items
        var response = await _client.ParseAsync<InvoiceV4>(inputSource);

        var prediction = response.Document.Inference.Prediction;

        return new InvoiceData
        {
            InvoiceNumber = prediction.InvoiceNumber?.Value,
            Date          = prediction.InvoiceDate?.Value,
            VendorName    = prediction.SupplierName?.Value,
            CustomerName  = prediction.CustomerName?.Value,
            Total         = prediction.TotalAmount?.Value,
            // Payment details transmitted to Mindee cloud
            PaymentDetails = prediction.SupplierPaymentDetails?
                .Select(pd => new PaymentDetail
                {
                    Iban          = pd.Iban,
                    Swift         = pd.Swift,
                    RoutingNumber = pd.RoutingNumber,
                    AccountNumber = pd.AccountNumber
                }).ToList()
        };
    }
}

using Mindee;
using Mindee.Input;
using Mindee.Product.Invoice;

public class MindeeInvoiceService
{
    private readonly MindeeClient _client;

    public MindeeInvoiceService(string apiKey)
    {
        // API key authenticates with Mindee cloud
        _client = new MindeeClient(apiKey);
    }

    public async Task<InvoiceData> ParseInvoiceAsync(string filePath)
    {
        // Document leaves your infrastructure at this line
        var inputSource = new LocalInputSource(filePath);

        // Full document transmitted to Mindee — bank details, tax IDs, line items
        var response = await _client.ParseAsync<InvoiceV4>(inputSource);

        var prediction = response.Document.Inference.Prediction;

        return new InvoiceData
        {
            InvoiceNumber = prediction.InvoiceNumber?.Value,
            Date          = prediction.InvoiceDate?.Value,
            VendorName    = prediction.SupplierName?.Value,
            CustomerName  = prediction.CustomerName?.Value,
            Total         = prediction.TotalAmount?.Value,
            // Payment details transmitted to Mindee cloud
            PaymentDetails = prediction.SupplierPaymentDetails?
                .Select(pd => new PaymentDetail
                {
                    Iban          = pd.Iban,
                    Swift         = pd.Swift,
                    RoutingNumber = pd.RoutingNumber,
                    AccountNumber = pd.AccountNumber
                }).ToList()
        };
    }
}

Imports Mindee
Imports Mindee.Input
Imports Mindee.Product.Invoice
Imports System.Threading.Tasks

Public Class MindeeInvoiceService
    Private ReadOnly _client As MindeeClient

    Public Sub New(apiKey As String)
        ' API key authenticates with Mindee cloud
        _client = New MindeeClient(apiKey)
    End Sub

    Public Async Function ParseInvoiceAsync(filePath As String) As Task(Of InvoiceData)
        ' Document leaves your infrastructure at this line
        Dim inputSource = New LocalInputSource(filePath)

        ' Full document transmitted to Mindee — bank details, tax IDs, line items
        Dim response = Await _client.ParseAsync(Of InvoiceV4)(inputSource)

        Dim prediction = response.Document.Inference.Prediction

        Return New InvoiceData With {
            .InvoiceNumber = prediction.InvoiceNumber?.Value,
            .Date = prediction.InvoiceDate?.Value,
            .VendorName = prediction.SupplierName?.Value,
            .CustomerName = prediction.CustomerName?.Value,
            .Total = prediction.TotalAmount?.Value,
            ' Payment details transmitted to Mindee cloud
            .PaymentDetails = prediction.SupplierPaymentDetails?.
                Select(Function(pd) New PaymentDetail With {
                    .Iban = pd.Iban,
                    .Swift = pd.Swift,
                    .RoutingNumber = pd.RoutingNumber,
                    .AccountNumber = pd.AccountNumber
                }).ToList()
        }
    End Function
End Class

The code is concise. The trade-off is that pd.Iban, pd.RoutingNumber, and pd.AccountNumber exist in the response only because Mindee received and processed a document containing them.

Detailed Feature Comparison

Mindee Approach

When your application calls _client.ParseAsync<InvoiceV4>(inputSource), the complete document file is uploaded to Mindee's cloud infrastructure. This is not metadata or a fingerprint — it is the full document content. The fields Mindee extracts from invoices are among the most sensitive data your organization handles:

// Every field in this response existed because Mindee received your document
var prediction = response.Document.Inference.Prediction;

// Financial identifiers — transmitted to Mindee cloud
var iban          = prediction.SupplierPaymentDetails?.FirstOrDefault()?.Iban;
var routingNumber = prediction.SupplierPaymentDetails?.FirstOrDefault()?.RoutingNumber;
var accountNumber = prediction.SupplierPaymentDetails?.FirstOrDefault()?.AccountNumber;
var swift         = prediction.SupplierPaymentDetails?.FirstOrDefault()?.Swift;

// Tax identification — transmitted to Mindee cloud
var vendorTaxId   = prediction.SupplierCompanyRegistrations?.FirstOrDefault()?.Value;
var customerReg   = prediction.CustomerCompanyRegistrations?.FirstOrDefault()?.Value;

// Business intelligence — transmitted to Mindee cloud
var lineItems = prediction.LineItems?
    .Select(li => new
    {
        li.Description,  // What you buy
        li.Quantity,     // How much you buy
        li.UnitPrice,    // What you pay per unit
        li.TotalAmount   // Line total
    }).ToList();

// Every field in this response existed because Mindee received your document
var prediction = response.Document.Inference.Prediction;

// Financial identifiers — transmitted to Mindee cloud
var iban          = prediction.SupplierPaymentDetails?.FirstOrDefault()?.Iban;
var routingNumber = prediction.SupplierPaymentDetails?.FirstOrDefault()?.RoutingNumber;
var accountNumber = prediction.SupplierPaymentDetails?.FirstOrDefault()?.AccountNumber;
var swift         = prediction.SupplierPaymentDetails?.FirstOrDefault()?.Swift;

// Tax identification — transmitted to Mindee cloud
var vendorTaxId   = prediction.SupplierCompanyRegistrations?.FirstOrDefault()?.Value;
var customerReg   = prediction.CustomerCompanyRegistrations?.FirstOrDefault()?.Value;

// Business intelligence — transmitted to Mindee cloud
var lineItems = prediction.LineItems?
    .Select(li => new
    {
        li.Description,  // What you buy
        li.Quantity,     // How much you buy
        li.UnitPrice,    // What you pay per unit
        li.TotalAmount   // Line total
    }).ToList();

' Every field in this response existed because Mindee received your document
Dim prediction = response.Document.Inference.Prediction

' Financial identifiers — transmitted to Mindee cloud
Dim iban = prediction.SupplierPaymentDetails?.FirstOrDefault()?.Iban
Dim routingNumber = prediction.SupplierPaymentDetails?.FirstOrDefault()?.RoutingNumber
Dim accountNumber = prediction.SupplierPaymentDetails?.FirstOrDefault()?.AccountNumber
Dim swift = prediction.SupplierPaymentDetails?.FirstOrDefault()?.Swift

' Tax identification — transmitted to Mindee cloud
Dim vendorTaxId = prediction.SupplierCompanyRegistrations?.FirstOrDefault()?.Value
Dim customerReg = prediction.CustomerCompanyRegistrations?.FirstOrDefault()?.Value

' Business intelligence — transmitted to Mindee cloud
Dim lineItems = prediction.LineItems? _
    .Select(Function(li) New With {
        li.Description,  ' What you buy
        li.Quantity,     ' How much you buy
        li.UnitPrice,    ' What you pay per unit
        li.TotalAmount   ' Line total
    }).ToList()

Mindee holds SOC 2 Type II certification and claims GDPR compliance. Compliance documentation establishes that Mindee follows security practices. It does not mean your data stays within your security perimeter. The document physically leaves your infrastructure, traverses the public internet, and is processed on hardware you do not own or control. For the free plan, document retention runs up to 30 days.

The problem is structural. If your use case involves processing customer-submitted invoices, you are transmitting your customers' vendor relationships, bank details, and purchasing patterns to a third party. If you operate under GLBA, HIPAA, CMMC, FedRAMP, or data residency regulations, cloud document processing through a third-party API requires explicit processor agreements and may be outright prohibited.

IronOCR Approach

IronOCR processes documents with zero external transmission. The OCR engine, language models, and preprocessing pipeline all run inside your process on your hardware. No network call is made at recognition time:

using IronOcr;
using System.Text.RegularExpressions;

public class LocalInvoiceExtractor
{
    private readonly IronTesseract _ocr = new IronTesseract();

    public InvoiceData ExtractInvoice(string filePath)
    {
        // Processing on your machine — no data transmission occurs
        var result = _ocr.Read(filePath);
        var text   = result.Text;

        return new InvoiceData
        {
            InvoiceNumber = ExtractPattern(text, @"Invoice\s*#?\s*:?\s*([A-Z0-9]+-?\d+)"),
            Date          = ExtractDate(text),
            VendorName    = ExtractVendorName(result),
            CustomerName  = ExtractCustomerName(text),
            Total         = ExtractAmount(text, @"Total\s*(?:Due)?\s*:?\s*\$?([\d,]+\.?\d*)"),
            Tax           = ExtractAmount(text, @"Tax\s*:?\s*\$?([\d,]+\.?\d*)")
        };
    }

    private string ExtractVendorName(OcrResult result)
    {
        // Vendor name typically appears in the top 15% of the invoice
        var maxY     = result.Lines.Max(l => l.Y + l.Height);
        var topLines = result.Lines
            .Where(l => l.Y < maxY * 0.15)
            .OrderBy(l => l.Y);
        return topLines.FirstOrDefault(l => l.Text.Length > 5)?.Text;
    }

    private string ExtractPattern(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        return match.Success ? match.Groups[1].Value : null;
    }
}

using IronOcr;
using System.Text.RegularExpressions;

public class LocalInvoiceExtractor
{
    private readonly IronTesseract _ocr = new IronTesseract();

    public InvoiceData ExtractInvoice(string filePath)
    {
        // Processing on your machine — no data transmission occurs
        var result = _ocr.Read(filePath);
        var text   = result.Text;

        return new InvoiceData
        {
            InvoiceNumber = ExtractPattern(text, @"Invoice\s*#?\s*:?\s*([A-Z0-9]+-?\d+)"),
            Date          = ExtractDate(text),
            VendorName    = ExtractVendorName(result),
            CustomerName  = ExtractCustomerName(text),
            Total         = ExtractAmount(text, @"Total\s*(?:Due)?\s*:?\s*\$?([\d,]+\.?\d*)"),
            Tax           = ExtractAmount(text, @"Tax\s*:?\s*\$?([\d,]+\.?\d*)")
        };
    }

    private string ExtractVendorName(OcrResult result)
    {
        // Vendor name typically appears in the top 15% of the invoice
        var maxY     = result.Lines.Max(l => l.Y + l.Height);
        var topLines = result.Lines
            .Where(l => l.Y < maxY * 0.15)
            .OrderBy(l => l.Y);
        return topLines.FirstOrDefault(l => l.Text.Length > 5)?.Text;
    }

    private string ExtractPattern(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        return match.Success ? match.Groups[1].Value : null;
    }
}

Imports IronOcr
Imports System.Text.RegularExpressions

Public Class LocalInvoiceExtractor
    Private ReadOnly _ocr As New IronTesseract()

    Public Function ExtractInvoice(filePath As String) As InvoiceData
        ' Processing on your machine — no data transmission occurs
        Dim result = _ocr.Read(filePath)
        Dim text = result.Text

        Return New InvoiceData With {
            .InvoiceNumber = ExtractPattern(text, "Invoice\s*#?\s*:?\s*([A-Z0-9]+-?\d+)"),
            .Date = ExtractDate(text),
            .VendorName = ExtractVendorName(result),
            .CustomerName = ExtractCustomerName(text),
            .Total = ExtractAmount(text, "Total\s*(?:Due)?\s*:?\s*\$?([\d,]+\.?\d*)"),
            .Tax = ExtractAmount(text, "Tax\s*:?\s*\$?([\d,]+\.?\d*)")
        }
    End Function

    Private Function ExtractVendorName(result As OcrResult) As String
        ' Vendor name typically appears in the top 15% of the invoice
        Dim maxY = result.Lines.Max(Function(l) l.Y + l.Height)
        Dim topLines = result.Lines _
            .Where(Function(l) l.Y < maxY * 0.15) _
            .OrderBy(Function(l) l.Y)
        Return topLines.FirstOrDefault(Function(l) l.Text.Length > 5)?.Text
    End Function

    Private Function ExtractPattern(text As String, pattern As String) As String
        Dim match = Regex.Match(text, pattern, RegexOptions.IgnoreCase)
        Return If(match.Success, match.Groups(1).Value, Nothing)
    End Function
End Class

Bank account numbers in the document never reach any system outside your infrastructure. Compliance scope covers your organization only. The IronOCR invoice OCR tutorial and the receipt scanning guide cover production-ready extraction patterns in detail.

The trade-off is explicit: Mindee's pre-trained models extract structured fields with no pattern-writing on your part. IronOCR requires building and maintaining extraction patterns. For most invoice formats, those patterns take a few hours to write and are straightforward to maintain. The data sovereignty benefit is permanent.

Mindee Approach

Mindee's supported document types are fixed at the API level. The pre-built APIs cover InvoiceV4, ReceiptV5, PassportV1, UsDriverLicenseV1, BankAccountDetailsV2, FinancialDocumentV1, and a small number of others. If your application needs to process a document type outside this list — contracts, medical forms, shipping labels, lease agreements, custom business forms — the options are custom model training on Mindee's Enterprise plan or finding a different solution.

// These work with Mindee pre-built APIs:
var invoiceResponse  = await _client.ParseAsync<InvoiceV4>(inputSource);
var receiptResponse  = await _client.ParseAsync<ReceiptV5>(inputSource);
var passportResponse = await _client.ParseAsync<PassportV1>(inputSource);

// These require custom training (Enterprise plan + labeling + lead time):
// client.ParseAsync<ContractV1>(inputSource);      // not available
// client.ParseAsync<MedicalFormV1>(inputSource);   // not available
// client.ParseAsync<ShippingLabelV1>(inputSource); // not available

// These work with Mindee pre-built APIs:
var invoiceResponse  = await _client.ParseAsync<InvoiceV4>(inputSource);
var receiptResponse  = await _client.ParseAsync<ReceiptV5>(inputSource);
var passportResponse = await _client.ParseAsync<PassportV1>(inputSource);

// These require custom training (Enterprise plan + labeling + lead time):
// client.ParseAsync<ContractV1>(inputSource);      // not available
// client.ParseAsync<MedicalFormV1>(inputSource);   // not available
// client.ParseAsync<ShippingLabelV1>(inputSource); // not available

Custom model training on Mindee requires sample documents, field labeling, training time, and Enterprise pricing. If your document processing requirements expand over time, each new document type is a separate engineering investment gated behind Mindee's training pipeline.

IronOCR Approach

IronOCR processes any document type using the same API and the same installation. Extraction logic is pattern-based code you write once and own entirely:

using IronOcr;
using System.Text.RegularExpressions;

public class UniversalDocumentProcessor
{
    private readonly IronTesseract _ocr = new IronTesseract();

    // Invoices — same approach as Mindee covers
    public InvoiceData ProcessInvoice(string file)
    {
        var result = _ocr.Read(file);
        return new InvoiceData
        {
            InvoiceNumber = ExtractPattern(result.Text, @"Invoice\s*#?\s*(\w+\d+)"),
            Total         = ExtractCurrency(result.Text, @"Total\s*:?\s*\$?([\d,]+\.?\d*)")
        };
    }

    // Contracts — Mindee has no pre-built API for these
    public ContractData ProcessContract(string file)
    {
        var result = _ocr.Read(file);
        return new ContractData
        {
            PartyA         = ExtractPattern(result.Text, @"between\s+(.+?)\s+and"),
            PartyB         = ExtractPattern(result.Text, @"and\s+(.+?)\s*\("),
            EffectiveDate  = ExtractDate(result.Text, @"Effective\s+Date\s*:?\s*(.+)"),
            ContractValue  = ExtractCurrency(result.Text, @"Value\s*:?\s*\$?([\d,]+)")
        };
    }

    // Medical forms — Mindee has no pre-built API for these
    public MedicalFormData ProcessMedicalForm(string file)
    {
        var result = _ocr.Read(file);
        return new MedicalFormData
        {
            PatientName = ExtractPattern(result.Text, @"Patient\s*Name\s*:?\s*(.+)"),
            MRN         = ExtractPattern(result.Text, @"MRN\s*:?\s*(\w+)"),
            Diagnosis   = ExtractPattern(result.Text, @"Diagnosis\s*:?\s*(.+)")
        };
    }

    private string ExtractPattern(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        return match.Success ? match.Groups[1].Value.Trim() : null;
    }

    private decimal? ExtractCurrency(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        if (match.Success && decimal.TryParse(
            match.Groups[1].Value.Replace(",", ""), out var value))
            return value;
        return null;
    }

    private DateTime? ExtractDate(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        if (match.Success && DateTime.TryParse(match.Groups[1].Value, out var date))
            return date;
        return null;
    }
}

using IronOcr;
using System.Text.RegularExpressions;

public class UniversalDocumentProcessor
{
    private readonly IronTesseract _ocr = new IronTesseract();

    // Invoices — same approach as Mindee covers
    public InvoiceData ProcessInvoice(string file)
    {
        var result = _ocr.Read(file);
        return new InvoiceData
        {
            InvoiceNumber = ExtractPattern(result.Text, @"Invoice\s*#?\s*(\w+\d+)"),
            Total         = ExtractCurrency(result.Text, @"Total\s*:?\s*\$?([\d,]+\.?\d*)")
        };
    }

    // Contracts — Mindee has no pre-built API for these
    public ContractData ProcessContract(string file)
    {
        var result = _ocr.Read(file);
        return new ContractData
        {
            PartyA         = ExtractPattern(result.Text, @"between\s+(.+?)\s+and"),
            PartyB         = ExtractPattern(result.Text, @"and\s+(.+?)\s*\("),
            EffectiveDate  = ExtractDate(result.Text, @"Effective\s+Date\s*:?\s*(.+)"),
            ContractValue  = ExtractCurrency(result.Text, @"Value\s*:?\s*\$?([\d,]+)")
        };
    }

    // Medical forms — Mindee has no pre-built API for these
    public MedicalFormData ProcessMedicalForm(string file)
    {
        var result = _ocr.Read(file);
        return new MedicalFormData
        {
            PatientName = ExtractPattern(result.Text, @"Patient\s*Name\s*:?\s*(.+)"),
            MRN         = ExtractPattern(result.Text, @"MRN\s*:?\s*(\w+)"),
            Diagnosis   = ExtractPattern(result.Text, @"Diagnosis\s*:?\s*(.+)")
        };
    }

    private string ExtractPattern(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        return match.Success ? match.Groups[1].Value.Trim() : null;
    }

    private decimal? ExtractCurrency(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        if (match.Success && decimal.TryParse(
            match.Groups[1].Value.Replace(",", ""), out var value))
            return value;
        return null;
    }

    private DateTime? ExtractDate(string text, string pattern)
    {
        var match = Regex.Match(text, pattern, RegexOptions.IgnoreCase);
        if (match.Success && DateTime.TryParse(match.Groups[1].Value, out var date))
            return date;
        return null;
    }
}

Imports IronOcr
Imports System.Text.RegularExpressions

Public Class UniversalDocumentProcessor
    Private ReadOnly _ocr As New IronTesseract()

    ' Invoices — same approach as Mindee covers
    Public Function ProcessInvoice(file As String) As InvoiceData
        Dim result = _ocr.Read(file)
        Return New InvoiceData With {
            .InvoiceNumber = ExtractPattern(result.Text, "Invoice\s*#?\s*(\w+\d+)"),
            .Total = ExtractCurrency(result.Text, "Total\s*:?\s*\$?([\d,]+\.?\d*)")
        }
    End Function

    ' Contracts — Mindee has no pre-built API for these
    Public Function ProcessContract(file As String) As ContractData
        Dim result = _ocr.Read(file)
        Return New ContractData With {
            .PartyA = ExtractPattern(result.Text, "between\s+(.+?)\s+and"),
            .PartyB = ExtractPattern(result.Text, "and\s+(.+?)\s*\("),
            .EffectiveDate = ExtractDate(result.Text, "Effective\s+Date\s*:?\s*(.+)"),
            .ContractValue = ExtractCurrency(result.Text, "Value\s*:?\s*\$?([\d,]+)")
        }
    End Function

    ' Medical forms — Mindee has no pre-built API for these
    Public Function ProcessMedicalForm(file As String) As MedicalFormData
        Dim result = _ocr.Read(file)
        Return New MedicalFormData With {
            .PatientName = ExtractPattern(result.Text, "Patient\s*Name\s*:?\s*(.+)"),
            .MRN = ExtractPattern(result.Text, "MRN\s*:?\s*(\w+)"),
            .Diagnosis = ExtractPattern(result.Text, "Diagnosis\s*:?\s*(.+)")
        }
    End Function

    Private Function ExtractPattern(text As String, pattern As String) As String
        Dim match = Regex.Match(text, pattern, RegexOptions.IgnoreCase)
        Return If(match.Success, match.Groups(1).Value.Trim(), Nothing)
    End Function

    Private Function ExtractCurrency(text As String, pattern As String) As Decimal?
        Dim match = Regex.Match(text, pattern, RegexOptions.IgnoreCase)
        If match.Success AndAlso Decimal.TryParse(match.Groups(1).Value.Replace(",", ""), value) Then
            Return value
        End If
        Return Nothing
    End Function

    Private Function ExtractDate(text As String, pattern As String) As DateTime?
        Dim match = Regex.Match(text, pattern, RegexOptions.IgnoreCase)
        If match.Success AndAlso DateTime.TryParse(match.Groups(1).Value, date) Then
            Return date
        End If
        Return Nothing
    End Function
End Class

When a new document type is required, adding it means writing extraction patterns — typically a few hours of work. No vendor engagement, no training pipeline, no additional cost. The read-specific-document tutorial and region-based OCR guide cover techniques for zone-based extraction on structured forms.

Mindee Approach

Mindee pricing is per-page and ongoing. Contact Mindee for current plan rates. Pages beyond the included allotment incur overage charges. Multi-page PDFs count each page separately. Development and testing pages count against the monthly quota.

For a medium-size accounts payable operation, per-page costs compound over time and scale with volume. A year-end processing spike or a large batch of multi-page invoices can push costs above the included tier without warning.

Batch processing with Mindee also requires rate-limit management. Each document is an individual API call, and calls must be spaced to avoid throttling:

// Mindee batch: each document is an API call with associated cost and rate limit
public async Task<List<InvoiceData>> ProcessBatchAsync(IEnumerable<string> files)
{
    var results = new List<InvoiceData>();

    foreach (var file in files)
    {
        var inputSource = new LocalInputSource(file);
        var response    = await _client.ParseAsync<InvoiceV4>(inputSource);
        results.Add(MapResult(response.Document.Inference.Prediction));

        // Rate limit compliance — required, not optional
        await Task.Delay(100);
    }

    return results;
}

// Mindee batch: each document is an API call with associated cost and rate limit
public async Task<List<InvoiceData>> ProcessBatchAsync(IEnumerable<string> files)
{
    var results = new List<InvoiceData>();

    foreach (var file in files)
    {
        var inputSource = new LocalInputSource(file);
        var response    = await _client.ParseAsync<InvoiceV4>(inputSource);
        results.Add(MapResult(response.Document.Inference.Prediction));

        // Rate limit compliance — required, not optional
        await Task.Delay(100);
    }

    return results;
}

Imports System.Collections.Generic
Imports System.Threading.Tasks

' Mindee batch: each document is an API call with associated cost and rate limit
Public Async Function ProcessBatchAsync(files As IEnumerable(Of String)) As Task(Of List(Of InvoiceData))
    Dim results As New List(Of InvoiceData)()

    For Each file In files
        Dim inputSource = New LocalInputSource(file)
        Dim response = Await _client.ParseAsync(Of InvoiceV4)(inputSource)
        results.Add(MapResult(response.Document.Inference.Prediction))

        ' Rate limit compliance — required, not optional
        Await Task.Delay(100)
    Next

    Return results
End Function

IronOCR Approach

IronOCR licensing is perpetual and volume-unlimited. The $1,499 Plus license covers three developers and three projects with no per-document cost. Processing 3,000 invoices per month or 300,000 per month costs nothing additional. Batch processing uses Parallel.ForEach with no rate-limit concerns:

using IronOcr;

public class IronOcrBatchProcessor
{
    private readonly IronTesseract _ocr = new IronTesseract();

    // No rate limits, no per-document cost, no network latency
    public List<InvoiceData> ProcessBatch(IEnumerable<string> files)
    {
        var results = new List<InvoiceData>();

        Parallel.ForEach(files, file =>
        {
            var result    = _ocr.Read(file);
            var extracted = BuildInvoiceData(result);
            lock (results) { results.Add(extracted); }
        });

        return results;
    }

    private InvoiceData BuildInvoiceData(OcrResult result)
    {
        return new InvoiceData
        {
            InvoiceNumber = ExtractPattern(result.Text, @"Invoice\s*#?\s*(\w+\d+)"),
            Total         = ExtractTotal(result.Text)
        };
    }
}

using IronOcr;

public class IronOcrBatchProcessor
{
    private readonly IronTesseract _ocr = new IronTesseract();

    // No rate limits, no per-document cost, no network latency
    public List<InvoiceData> ProcessBatch(IEnumerable<string> files)
    {
        var results = new List<InvoiceData>();

        Parallel.ForEach(files, file =>
        {
            var result    = _ocr.Read(file);
            var extracted = BuildInvoiceData(result);
            lock (results) { results.Add(extracted); }
        });

        return results;
    }

    private InvoiceData BuildInvoiceData(OcrResult result)
    {
        return new InvoiceData
        {
            InvoiceNumber = ExtractPattern(result.Text, @"Invoice\s*#?\s*(\w+\d+)"),
            Total         = ExtractTotal(result.Text)
        };
    }
}

Imports IronOcr
Imports System.Collections.Generic
Imports System.Threading.Tasks

Public Class IronOcrBatchProcessor
    Private ReadOnly _ocr As New IronTesseract()

    ' No rate limits, no per-document cost, no network latency
    Public Function ProcessBatch(files As IEnumerable(Of String)) As List(Of InvoiceData)
        Dim results As New List(Of InvoiceData)()

        Parallel.ForEach(files, Sub(file)
                                    Dim result = _ocr.Read(file)
                                    Dim extracted = BuildInvoiceData(result)
                                    SyncLock results
                                        results.Add(extracted)
                                    End SyncLock
                                End Sub)

        Return results
    End Function

    Private Function BuildInvoiceData(result As OcrResult) As InvoiceData
        Return New InvoiceData With {
            .InvoiceNumber = ExtractPattern(result.Text, "Invoice\s*#?\s*(\w+\d+)"),
            .Total = ExtractTotal(result.Text)
        }
    End Function
End Class

Three-year costs for a high-volume operation: IronOCR Professional at $2,999 one-time versus Mindee's ongoing per-page fees. The break-even point is typically within the first few months of use. The IronOCR licensing page covers all tier details.

Mindee Approach

Mindee's async pattern is network I/O-driven. Every call is async because every call makes an HTTP request to Mindee's API endpoint. There is no synchronous path. In an application that processes documents at user request, the Mindee workflow requires async infrastructure throughout the calling stack:

// Every Mindee operation is async — network I/O has no synchronous equivalent
public async Task<InvoiceData> ProcessInvoiceAsync(string file)
{
    try
    {
        var inputSource = new LocalInputSource(file);
        var response    = await _client.ParseAsync<InvoiceV4>(inputSource);
        return MapResult(response.Document.Inference.Prediction);
    }
    catch (MindeeException ex)
    {
        // API errors, rate limits, authentication failures
        Console.WriteLine($"Mindee API error: {ex.Message}");
        throw;
    }
    catch (HttpRequestException ex)
    {
        // Network failure — processing halts entirely
        // No offline fallback path exists
        throw new Exception("Mindee requires internet connectivity", ex);
    }
}

// Every Mindee operation is async — network I/O has no synchronous equivalent
public async Task<InvoiceData> ProcessInvoiceAsync(string file)
{
    try
    {
        var inputSource = new LocalInputSource(file);
        var response    = await _client.ParseAsync<InvoiceV4>(inputSource);
        return MapResult(response.Document.Inference.Prediction);
    }
    catch (MindeeException ex)
    {
        // API errors, rate limits, authentication failures
        Console.WriteLine($"Mindee API error: {ex.Message}");
        throw;
    }
    catch (HttpRequestException ex)
    {
        // Network failure — processing halts entirely
        // No offline fallback path exists
        throw new Exception("Mindee requires internet connectivity", ex);
    }
}

Imports System
Imports System.Threading.Tasks

Public Async Function ProcessInvoiceAsync(file As String) As Task(Of InvoiceData)
    Try
        Dim inputSource = New LocalInputSource(file)
        Dim response = Await _client.ParseAsync(Of InvoiceV4)(inputSource)
        Return MapResult(response.Document.Inference.Prediction)
    Catch ex As MindeeException
        ' API errors, rate limits, authentication failures
        Console.WriteLine($"Mindee API error: {ex.Message}")
        Throw
    Catch ex As HttpRequestException
        ' Network failure — processing halts entirely
        ' No offline fallback path exists
        Throw New Exception("Mindee requires internet connectivity", ex)
    End Try
End Function

If the network is unavailable — a connectivity outage, a Mindee service disruption, a rate limit — processing stops. There is no degraded mode and no local fallback. Field deployments, air-gapped environments, and scenarios requiring guaranteed processing continuity are incompatible with Mindee's architecture.

IronOCR Approach

IronOCR is synchronous by default because local processing has no network I/O. For applications that require async interfaces for consistency with other I/O-bound operations, Task.Run wraps the synchronous call cleanly. The underlying processing is CPU-bound, not network-bound:

using IronOcr;

public class IronOcrInvoiceService
{
    private readonly IronTesseract _ocr = new IronTesseract();

    // Synchronous — local CPU-bound processing, no network dependency
    public InvoiceData ProcessInvoice(string file)
    {
        // Works with no internet, through any outage, in any environment
        var result = _ocr.Read(file);
        return BuildInvoiceData(result);
    }

    // Async wrapper for applications that require async consistency
    public async Task<InvoiceData> ProcessInvoiceAsync(string file)
    {
        return await Task.Run(() => ProcessInvoice(file));
    }
}

using IronOcr;

public class IronOcrInvoiceService
{
    private readonly IronTesseract _ocr = new IronTesseract();

    // Synchronous — local CPU-bound processing, no network dependency
    public InvoiceData ProcessInvoice(string file)
    {
        // Works with no internet, through any outage, in any environment
        var result = _ocr.Read(file);
        return BuildInvoiceData(result);
    }

    // Async wrapper for applications that require async consistency
    public async Task<InvoiceData> ProcessInvoiceAsync(string file)
    {
        return await Task.Run(() => ProcessInvoice(file));
    }
}

Imports IronOcr

Public Class IronOcrInvoiceService
    Private ReadOnly _ocr As New IronTesseract()

    ' Synchronous — local CPU-bound processing, no network dependency
    Public Function ProcessInvoice(file As String) As InvoiceData
        ' Works with no internet, through any outage, in any environment
        Dim result = _ocr.Read(file)
        Return BuildInvoiceData(result)
    End Function

    ' Async wrapper for applications that require async consistency
    Public Async Function ProcessInvoiceAsync(file As String) As Task(Of InvoiceData)
        Return Await Task.Run(Function() ProcessInvoice(file))
    End Function
End Class

IronOCR runs identically in Docker containers with no outbound access, in Azure Functions with egress restrictions, in government secure facilities, and in factory environments without internet connectivity. The Docker deployment guide and AWS deployment guide cover environment-specific configuration. For teams that genuinely want async processing with progress reporting, IronOCR's async OCR guide covers the native async API.

Compliance Requirements Emerge After Initial Deployment

Teams often start with Mindee for a quick proof of concept with non-sensitive documents, then discover that production requirements impose data sovereignty constraints. An accounts payable automation project that begins with internal test invoices encounters a different picture when real vendor invoices arrive carrying bank account numbers, EIN values, and pricing data that your vendor agreements classify as confidential. At that point, the choice is negotiating a DPA with Mindee and accepting ongoing cloud transmission risk, or switching to local processing. The migration is documented and achievable, but it requires rewriting extraction logic to replace Mindee's structured output with custom patterns against result.Text and result.Lines.

Volume Growth Makes Per-Page Pricing Untenable

A business processing 500 invoices per month stays comfortably within Mindee's entry plan. At higher volumes, per-page costs compound significantly, and large operations may require Enterprise pricing negotiations. Teams that correctly project volume growth often recalculate the break-even point before scaling: for a 5-developer team, IronOCR Professional at $2,999 one-time typically recovers its cost against ongoing Mindee fees within the first few months, with zero ongoing cost thereafter regardless of volume.

New Document Types Exceed Mindee's Pre-Built Catalog

The practical limit of Mindee's model surfaces when a business process requires document types outside invoices, receipts, and passports. Contracts, purchase orders with custom layouts, medical referral forms, lease agreements, and industry-specific forms have no Mindee pre-built API. Custom model training requires Enterprise pricing, sample document collection, labeling work, and training lead time. Teams managing multiple document types across a line-of-business application find that the marginal cost of adding each new Mindee model type makes the economics of a generalist local library compelling.

Air-Gapped or Restricted-Network Environments

Government contractors, defense subcontractors, financial institutions operating under strict network egress controls, and industrial automation systems deployed in facilities without reliable internet access cannot use Mindee as designed. The requirement is not unusual: many healthcare networks deliberately restrict outbound internet access for systems processing patient documents. IronOCR runs identically in restricted environments because it has no outbound dependencies at runtime.

Invoice Processing for Customer-Submitted Documents

When your application accepts documents from customers rather than processing your own internal invoices, the privacy calculation changes significantly. Transmitting a customer's invoice to Mindee means Mindee receives data about your customer's vendors, payment terms, and business relationships. Your customer submitted that document to your application, not to a third-party AI service. For SaaS platforms, fintech applications, and document management systems handling customer data, this distinction determines whether cloud document processing is appropriate regardless of Mindee's compliance certifications.

Building Extraction Patterns from Raw OCR Output

The core paradigm shift when moving from Mindee to IronOCR is that structured output must be built rather than received. Mindee returns prediction.InvoiceNumber?.Value as a pre-parsed field. IronOCR returns result.Text as a string, and your code applies regex patterns to extract fields.

using IronOcr;
using System.Text.RegularExpressions;

var ocr    = new IronTesseract();
var result = ocr.Read("invoice.jpg");

// Build the extraction logic Mindee previously handled server-side
var invoiceNumber = Regex.Match(result.Text,
    @"Invoice\s*#?\s*:?\s*([A-Z0-9]+-?\d+)",
    RegexOptions.IgnoreCase).Groups[1].Value;

var totalAmount = Regex.Match(result.Text,
    @"Total\s*(?:Due)?\s*:?\s*\$?([\d,]+\.?\d*)",
    RegexOptions.IgnoreCase).Groups[1].Value;

using IronOcr;
using System.Text.RegularExpressions;

var ocr    = new IronTesseract();
var result = ocr.Read("invoice.jpg");

// Build the extraction logic Mindee previously handled server-side
var invoiceNumber = Regex.Match(result.Text,
    @"Invoice\s*#?\s*:?\s*([A-Z0-9]+-?\d+)",
    RegexOptions.IgnoreCase).Groups[1].Value;

var totalAmount = Regex.Match(result.Text,
    @"Total\s*(?:Due)?\s*:?\s*\$?([\d,]+\.?\d*)",
    RegexOptions.IgnoreCase).Groups[1].Value;

Imports IronOcr
Imports System.Text.RegularExpressions

Dim ocr As New IronTesseract()
Dim result = ocr.Read("invoice.jpg")

' Build the extraction logic Mindee previously handled server-side
Dim invoiceNumber = Regex.Match(result.Text, _
    "Invoice\s*#?\s*:?\s*([A-Z0-9]+-?\d+)", _
    RegexOptions.IgnoreCase).Groups(1).Value

Dim totalAmount = Regex.Match(result.Text, _
    "Total\s*(?:Due)?\s*:?\s*\$?([\d,]+\.?\d*)", _
    RegexOptions.IgnoreCase).Groups(1).Value

Pattern libraries for common invoice formats are not complex — most invoice totals follow four or five predictable label patterns. Testing against a representative sample of 20-30 real invoices from your vendor set before deployment is sufficient to catch edge cases. The image quality correction guide covers preprocessing options that improve recognition accuracy on lower-quality scans.

Confidence Handling Changes

Mindee provides per-field confidence values: prediction.TotalAmount?.Confidence. IronOCR provides overall document confidence via result.Confidence, which represents the engine's aggregate certainty across the recognized text. For invoice processing, an overall confidence below 80% typically indicates a scan quality problem that benefits from preprocessing:

var ocr    = new IronTesseract();
var result = ocr.Read("invoice.jpg");

if (result.Confidence < 80)
{
    // Low confidence — apply preprocessing and retry
    using var input = new OcrInput();
    input.LoadImage("invoice.jpg");
    input.Deskew();
    input.DeNoise();
    input.EnhanceResolution(300);
    result = ocr.Read(input);
}

Console.WriteLine($"Confidence: {result.Confidence}%");

var ocr    = new IronTesseract();
var result = ocr.Read("invoice.jpg");

if (result.Confidence < 80)
{
    // Low confidence — apply preprocessing and retry
    using var input = new OcrInput();
    input.LoadImage("invoice.jpg");
    input.Deskew();
    input.DeNoise();
    input.EnhanceResolution(300);
    result = ocr.Read(input);
}

Console.WriteLine($"Confidence: {result.Confidence}%");

Imports IronOcr

Dim ocr As New IronTesseract()
Dim result = ocr.Read("invoice.jpg")

If result.Confidence < 80 Then
    ' Low confidence — apply preprocessing and retry
    Using input As New OcrInput()
        input.LoadImage("invoice.jpg")
        input.Deskew()
        input.DeNoise()
        input.EnhanceResolution(300)
        result = ocr.Read(input)
    End Using
End If

Console.WriteLine($"Confidence: {result.Confidence}%")

The confidence score documentation explains confidence interpretation and thresholds.

Async to Sync Pattern Adjustment

Mindee's entire API surface is async because it is network I/O. IronOCR's core methods are synchronous. If your existing codebase uses Mindee through async methods and needs to maintain an async interface for architectural consistency, wrap the synchronous IronOCR call:

// Existing async interface preserved
public async Task<InvoiceData> ParseInvoiceAsync(string filePath)
{
    // Task.Run offloads CPU-bound work from the calling thread
    return await Task.Run(() =>
    {
        var ocr    = new IronTesseract();
        var result = ocr.Read(filePath);
        return BuildInvoiceData(result);
    });
}

// Existing async interface preserved
public async Task<InvoiceData> ParseInvoiceAsync(string filePath)
{
    // Task.Run offloads CPU-bound work from the calling thread
    return await Task.Run(() =>
    {
        var ocr    = new IronTesseract();
        var result = ocr.Read(filePath);
        return BuildInvoiceData(result);
    });
}

Imports System.Threading.Tasks

' Existing async interface preserved
Public Async Function ParseInvoiceAsync(filePath As String) As Task(Of InvoiceData)
    ' Task.Run offloads CPU-bound work from the calling thread
    Return Await Task.Run(Function()
                              Dim ocr As New IronTesseract()
                              Dim result = ocr.Read(filePath)
                              Return BuildInvoiceData(result)
                          End Function)
End Function

For high-throughput ASP.NET scenarios, the IronOCR async guide covers thread pool considerations and the speed optimization guide addresses throughput tuning for batch workloads.

Removing Cloud Infrastructure Dependencies

Moving from Mindee removes the network egress requirements, API key rotation procedures, and Mindee service availability from your operational runbook. The Mindee API key stored in configuration is removed. Network egress rules permitting outbound HTTPS to Mindee endpoints can be revoked. Retry logic wrapping HttpRequestException is no longer needed. Operational complexity shrinks to the IronOCR license key set once at application startup: IronOcr.License.LicenseKey = Configuration["IronOCR:LicenseKey"].