drckarim/swiftocrkit

Features

Optical Character Recognition (OCR)

• Extract text from UIImage, NSImage, or CGImage
• Async/await API
• Configurable recognition levels (.accurate or .fast)
• Language correction support
• Custom language selection
• Minimum text height filtering
• Comprehensive error handling

Requirements

• iOS 15.0+
• macOS 12.0+
• Swift 5.7+
• Xcode 14.0+

Installation

Swift Package Manager

Add SwiftOCRKit using Xcode:

1. File → Add Packages
2. Enter the repository URL:

`` https://github.com/DrcKarim/SwiftOCRKit ``

1. Add the package to your target

Or add it to your Package.swift:

dependencies: [
    .package(url: "https://github.com/DrcKarim/SwiftOCRKit", from: "1.0.0")
]

Usage

Import the library:

import SwiftOCRKit

Basic OCR Example

// iOS
let image = UIImage(named: "document")!
let text = try await VisionOCR.recognizeText(from: image)
print(text)

// macOS
let image = NSImage(named: "document")!
let text = try await VisionOCR.recognizeText(from: image)
print(text)

Advanced OCR with Configuration

let config = VisionOCRConfiguration(
    recognitionLevel: .accurate,        // or .fast for speed
    usesLanguageCorrection: true,       // improves accuracy
    minimumTextHeight: 0.02,            // filter small text
    recognitionLanguages: ["en-US"]     // specify languages
)

let text = try await VisionOCR.recognizeText(from: image, configuration: config)
print(text)

Error Handling

do {
    let text = try await VisionOCR.recognizeText(from: image)
    print("Recognized text:", text)
} catch VisionOCRError.invalidImage {
    print("The provided image is invalid")
} catch VisionOCRError.noTextFound {
    print("No text was detected")
} catch VisionOCRError.visionError(let error) {
    print("Vision error:", error.localizedDescription)
} catch {
    print("Unexpected error:", error)
}

Supported Platforms

| Platform | Supported | |----------|-----------| | iOS | Yes | | macOS | Yes | | watchOS | No | | tvOS | No |

Configuration Options

| Option | Type | Default | Description | |--------|------|---------|-------------| | recognitionLevel | .accurate or .fast | .accurate | Balance between speed and accuracy | | usesLanguageCorrection | Bool | true | Apply language correction for better accuracy | | minimumTextHeight | Float | 0.0 | Minimum text height (0.0 to 1.0). Filters out small text | | recognitionLanguages | [String] | ` | Specific languages to recognize (e.g., ["en-US", "fr-FR"]`) |

API Reference

VisionOCR

// Basic API (uses default configuration)
static func recognizeText(from image: UIImage) async throws -> String
static func recognizeText(from image: NSImage) async throws -> String
static func recognizeText(from cgImage: CGImage) async throws -> String

// Advanced API (custom configuration)
static func recognizeText(from image: UIImage, configuration: VisionOCRConfiguration) async throws -> String
static func recognizeText(from image: NSImage, configuration: VisionOCRConfiguration) async throws -> String
static func recognizeText(from cgImage: CGImage, configuration: VisionOCRConfiguration) async throws -> String

VisionOCRError

enum VisionOCRError: Error {
    case invalidImage           // Image could not be processed
    case noTextFound           // No text detected in image
    case visionError(Error)    // Underlying Vision framework error
}

Architecture

SwiftOCRKit follows a clean, modular structure:

SwiftOCRKit/
└── OCR/
    ├── VisionOCR.swift               // Main OCR implementation
    ├── VisionOCRConfiguration.swift  // Configuration options
    └── VisionOCRError.swift          // Error types

Why SwiftOCRKit?

Apple Vision is powerful but verbose. SwiftOCRKit simplifies it into a clean, modern API.

Before (vanilla Vision):

let request = VNRecognizeTextRequest { request, error in
    guard error == nil else { return }
    let observations = request.results as? [VNRecognizedTextObservation] ?? []
    let text = observations.compactMap { $0.topCandidates(1).first?.string }.joined(separator: "\n")
    // handle callback-based result...
}
request.recognitionLevel = .accurate
request.usesLanguageCorrection = true
let handler = VNImageRequestHandler(cgImage: cgImage)
try handler.perform([request])

After (SwiftOCRKit):

let text = try await VisionOCR.recognizeText(from: image)

Roadmap

• [ ] Batch image processing
• [ ] PDF text extraction
• [ ] Confidence scores per text block
• [ ] Text region detection (bounding boxes)
• [ ] Barcode and QR code scanning
• [ ] Document detection and perspective correction

Performance Tips

1. Use .fast recognition level for real-time or video processing
2. Use .accurate for document scanning and high-quality text
3. Set minimumTextHeight to filter out noise and small artifacts
4. Specify recognitionLanguages when you know the language for better accuracy

Contributing

Contributions are welcome!

You can:

• Open issues for bugs or feature requests
• Submit pull requests with improvements
• Suggest new Vision-based features

Please keep changes focused and consistent with the existing API style.

License

MIT License Copyright (c) 2026 Made with ❤️ by Karim Bouchaane

Package Metadata

Repository: drckarim/swiftocrkit

Default branch: main

README: README.md