We integrate camera-based credit card scanning into mobile apps. Manual card entry causes errors and abandonment; our solutions cut input time to 1-2 seconds and reduce cart abandonment by 15-20%. For a $50K monthly revenue app, that can save over $10,000 per month. Integration starts from $500, and typical savings exceed $10,000 monthly. Our team has 5+ years of mobile development experience and 30+ successful payment integration projects.
What are the best libraries for card scanning?
| Library | Support | Accuracy | Complexity | Cost |
|---|---|---|---|---|
| Card.io (PayPal) | No longer maintained | Medium | Low | Free |
| Stripe CardScan / Bouncer | Active, Apache 2.0 | >95% | Medium | Free |
| Vision (iOS) + ML Kit (Android) | Apple/Google | ~90% (needs filtering) | High | Free |
Card.io was historically popular but is no longer maintained — on modern iOS versions there are AVCaptureSession crashes. We don't recommend it for new projects. Stripe CardScan (formerly Bouncer) is a neural network solution on TFLite/CoreML, works offline. On iOS, integrate via SPM: https://github.com/stripe/stripe-ios, on Android via Gradle: com.stripe:stripecardscan. Its code is open, allowing adaptation for specific requirements. Native OCR frameworks (VNRecognizeTextRequest, ML Kit Text Recognition v2) require no third-party dependencies, but you need to implement card region detection and result stabilization (confirmation via 3-5 consecutive identical results).
How to ensure recognition accuracy?
Recognition accuracy depends on shooting conditions. In good lighting, Stripe CardScan achieves 97% correct numbers — 20% more accurate than native solutions in low light. In low light or glare, accuracy drops to 85-90%. We recommend implementing Luhn check and frame illumination in low light. For result stabilization, use confirmation via 3 consecutive matches.
| Condition | Stripe CardScan | Vision/ML Kit |
|---|---|---|
| Good lighting | 97% | 92% |
| Low light | 85% | 78% |
| Glare | 80% | 72% |
Native frameworks vs. ready SDKs
A ready SDK (Stripe CardScan) provides high accuracy out of the box and reduces development time. But if you need full UI customization or work in specific conditions (non-standard cards, poor lighting) — native Vision/ML Kit give you full control. We combine approaches: for MVP we use Stripe CardScan, for production with unique requirements we write our own module on native APIs.
Recognizable card data
Card number is the primary target, accuracy >95%. Expiry date (MM/YY) is recognized confidently. Cardholder name — OCR handles it worse due to font variability and embossing. We recommend: number + expiry scanned automatically, name manually filled as optional. Never recognize or save CVV/CVC — it's on the back, and storing it violates PCI DSS requirements. This is not just a rule, but a mandatory condition for passing security audit.
Implementation on Flutter
The card_scanner package on pub.dev works via Platform Channels to native SDKs. Alternative: google_mlkit_text_recognition with custom parser. For production apps, Stripe CardScan via FFI is preferable: more accurate and reliable in non-standard conditions (poor lighting, non-standard card design). Flutter developers should consider that some stabilization methods require native code — we help with the wrapper.
Permissions and privacy for App Store review
iOS: NSCameraUsageDescription in Info.plist with wording "for scanning card during payment". Android: uses-permission android:name="android.permission.CAMERA" + runtime request via ActivityResultContracts.RequestPermission. Camera frames are processed only in memory — not saved to disk, not sent to server. This is critical for passing App Store and Play Market review. For more details, see Stripe CardScan documentation.
Scope of work (turnkey solution)
- Requirements analysis and optimal stack selection (Bouncer, Vision/ML Kit, Flutter package).
- SDK integration with custom UI (camera preview, card frame, animation).
- Implementation of stabilization logic and false positive filtering.
- Testing on different devices and in low-light conditions.
- Integration documentation and support for store publication.
Timelines and collaboration process
Integration of Stripe CardScan with basic UI: 1-2 days. Custom UI + native Vision/ML Kit with own stabilization logic: 2-3 days. Full cycle — from analysis to deployment — takes 3 to 5 days depending on complexity. Cost is calculated individually, starting from $500. We offer turnkey integration in 1-2 days. Write to us for a free project assessment — we'll analyze requirements and offer the best solution.
In-app card scanning with our solution is 3 times faster than manual entry. With a credit card scanning camera, users can capture details in seconds. We provide mobile OCR for cards that is PCI DSS-compliant. Our services cover credit card number recognition, card scanning on iOS and Android, and we are your trusted Card.io alternative. Want to speed up payments and reduce declines? Contact us to discuss details — we'll analyze requirements and propose the best option.