Native, on-device AI image super-resolution and restoration for iOS 18+ / iPadOS 18+. Built with Swift 6 and Apple Core ML.
- On-Device Inference: Hardware-accelerated execution on Apple Neural Engine (ANE), Metal GPU, and CPU.
- Super-Resolution Algorithms: 27 models across 9 architectures, including Real-CUGAN, Real-ESRGAN, Waifu2x, SRMD, BSRGAN, Real-ESRNet, and Manga Clean.
- Tiling Engine: Overlap decomposition and seamless stitching to process multi-megapixel images without memory pressure.
- Interactive Comparison: Full-screen before-and-after slider with synchronized pan and pinch-zoom.
- Batch Processing: Multi-image queue for automated background upscaling.
- Storage & Offline Management: Built-in model manager with on-demand downloads and swipe-to-delete.
- Privacy by Design: 100% offline, zero network requests during inference, zero analytics.
# Build unsigned IPA
./build.sh
# Clean build
./build.sh --cleanOutputs .ipa and .app to build_output/.
- iOS 18.0+ / iPadOS 18.0+
- Apple Silicon (A12 Bionic or newer recommended)
MIT