Vision & VisionKit (iOS)

On-device computer vision for a fintech app: OCR of cards/cheques/IDs, barcode/QR scanning, document capture for KYC, and custom Core ML inference. Everything here runs on device — no image ever leaves the phone, which is the right posture for KYC/PII. Grounded in Apple's Vision and VisionKit frameworks.

• Two Vision APIs coexist. The modern Swift API (RecognizeTextRequest, struct/class requests with async .perform(on:)) needs iOS 18+. The legacy API (VNRecognizeTextRequest + VNImageRequestHandler) works back to iOS 13 and is still the path when you support iOS 17 or earlier. Same ML models underneath; different surface.
• VisionKit is a separate framework for UI: full-screen live scanners and Live Text overlays. Reach for it before hand-rolling an AVCaptureSession + Vision loop.
• Do NOT confuse with visionOS. This skill is Apple's image-analysis Vision, unrelated to the Apple Vision Pro platform.

Contents