Google ML Kit
Google’s on-device machine learning SDK for mobile. Cross-platform (Android + iOS). Text Recognition API (v2, graduated) performs OCR locally with no network.
Why It Matters for Project Aries
If Project Aries ever targets Android or needs cross-platform OCR consistency, ML Kit is the primary option. On iOS, it competes with Apple Vision — offering 6x faster processing at the cost of fewer API features and Latin-only script support.
Text Recognition API
- Single processing mode (no fast/accurate distinction)
- Returns text blocks with bounding boxes
- Additional language detection API
- No confidence scores, custom words, or language correction
Performance vs Apple Vision
Benchmarked on iPhone 12 (2021):
| Metric | ML Kit | Apple Vision |
|---|---|---|
| Average speed | ~0.05s | ~0.31s |
| Low-resolution text | Slightly better | Slightly worse |
| Rotated text (>20°) | Slightly worse | Slightly better |
| Script support | Latin only | Latin + Chinese |
ML Kit is 6x faster, making it the clear choice for real-time camera preview OCR.
Platform
iOS (CocoaPods/SPM) and Android. Minimum Xcode 14.2 for iOS.
Related Pages
- apple-vision-framework — Competing on-device OCR framework (richer API, slower)
- ios-receipt-scanning — Comprehensive receipt scanning guide