sorty_kyc_capture

A Flutter package that provides the full UX flow for KYC (Know Your Customer) identity verification — including document capture, liveness selfie detection, and automatic image compression — ready to integrate with AWS Rekognition or any face-matching backend.

Features

• Document capture — Full-screen camera with an ID-card shaped overlay (ISO/IEC 7810 ratio) and alignment guides.
• Liveness selfie — Front camera with real-time face detection via google_mlkit_face_detection. Capture is only enabled when exactly one face is detected AND the user smiles (probability > 0.7) or blinks, preventing photo spoofing.
• Auto image compression — All captured images are compressed to JPEG ≤ 800 KB using flutter_image_compress, ideal for S3 uploads or REST APIs.
• Permission handling — Built-in camera permission gate widget with retry and system settings fallback.
• Structured result — Returns a KycResult(idFront, idBack, selfie) with all three compressed File objects.

Installation

dependencies:
  sorty_kyc_capture: ^0.1.0

Native setup

Android — AndroidManifest.xml

<uses-permission android:name="android.permission.CAMERA" />

Set minSdkVersion 21 in android/app/build.gradle.

iOS — Info.plist

<key>NSCameraUsageDescription</key>
<string>Required for identity verification (KYC).</string>

Set platform :ios, '14.0' in your Podfile.

Usage

import 'package:sorty_kyc_capture/sorty_kyc_capture.dart';

Navigator.of(context).push(MaterialPageRoute(
  builder: (_) => KycFlowScreen(
    onComplete: (KycResult result) async {
      // All files are JPEG, ≤ 800 KB, EXIF stripped
      print(result.idFront.path);  // compressed ID front
      print(result.idBack.path);   // compressed ID back
      print(result.selfie.path);   // liveness-verified selfie

      // Upload to your backend / S3, then:
      await result.dispose(); // deletes temp files
    },
    onCancel: () => Navigator.of(context).pop(),
  ),
));

Flow

KycFlowScreen
  ├── Step 1 → IdCaptureScreen  ("Frente del documento")
  ├── Step 2 → IdCaptureScreen  ("Reverso del documento")
  └── Step 3 → SelfieCaptureScreen  (liveness required)
                    ↓
               KycResult(idFront, idBack, selfie)

Liveness detection logic

Signal	Threshold	Purpose
smilingProbability	> 0.70	Active liveness (user must act)
leftEyeOpenProbability or rightEyeOpenProbability	< 0.25	Passive liveness (blink detection)
Face count	== 1	Prevents multi-face bypass

Processing is throttled to one frame every 150 ms to avoid overloading mid-range devices.

AWS Rekognition integration

Once you receive the KycResult, your backend should:

1. Upload the three files to a temporary S3 bucket.
2. Call rekognition:CompareFaces(source=selfie, target=idFront) — validate Similarity >= 95.
3. Call rekognition:DetectText(image=idFront) to extract document data (name, ID number, DOB).
4. Delete the S3 objects after verification.

Dependencies

Package	Purpose
camera	Full camera hardware control
google_mlkit_face_detection	On-device face & liveness detection
flutter_image_compress	JPEG compression
permission_handler	Runtime camera permission
path_provider	Temp directory for output files

License

MIT