A lightweight, fast, AI-powered image viewer for macOS.
Small & Fast
- Native macOS AppKit application — no Electron, no web technologies
- Minimal memory footprint, instant launch
- Smooth image transitions and responsive gestures
AI-Powered
- One-click AI auto-enhance (dedup + watermark removal)
- AI super-resolution (Real-ESRGAN)
- AI watermark removal (U2Net)
- Auto AI enhance on image load
Image Viewing
- Support for JPEG, PNG, HEIC, HEIF, GIF, WebP, BMP, TIFF
- Live Photo playback (HEIC + MOV pairs)
- Zoom, pan, rotate, crop
- Slideshow mode
- Duplicate detection and comparison
Keyboard-First
- Full keyboard navigation
- Customizable shortcuts
- Quick actions at your fingertips
Download from GitHub Releases:
https://github.com/J-Liu/PixAI/releases
Since PixAI is not notarized by Apple, you need to bypass the security check to run it.
Method 1: Command Line
xattr -cr /path/to/PixAI.appMethod 2: System Preferences
- Right-click
PixAI.appand select "Open" - Click "Open" in the security dialog
- Or go to System Preferences → Privacy & Security → click "Open Anyway"
- macOS 14.0 (Sonoma) or later
- Apple Silicon (M1/M2/M3/M4) or Intel Mac
PixAI uses local AI models that run entirely on your Mac — no cloud, no internet required for inference.
Models are downloaded on first use (or on startup if enabled in Preferences):
- Real-ESRGAN — AI super-resolution for upscaling images
- U2Net — AI watermark detection and removal
Settings are stored in ~/.pixai/config.json.
Open Preferences with Cmd+, to configure:
- General: language, transition animations
- Slideshow: interval timing
- AI: auto-enhance options, one-click mode
- Advanced: proxy settings, image cache, logging
PixAI can play Apple Live Photos exported from iPhone. On macOS, a Live Photo consists of two files:
IMG_1234.HEIC— the still imageIMG_1234.MOV— the 3-second video
Requirements for playback:
- Both files must be in the same directory
- Both files must have the same base name (e.g.,
IMG_1234.HEICandIMG_1234.MOV) - The MOV file must contain Live Photo metadata
Exporting from Photos app:
- Select the Live Photo in Photos
- File → Export → Export Unmodified Original
- Both HEIC and MOV files will be exported
When you open a Live Photo HEIC file in PixAI, it will automatically detect the companion MOV and show playback controls:
- Live Photo badge (bottom-left) — click to play/stop
- Mute button — toggle audio on/off
Auto-play can be enabled/disabled in Preferences.
PixAI supports:
- English
- 简体中文 (Simplified Chinese)
- 繁體中文 (Traditional Chinese)
Change language in Preferences → General.
Built with Swift and AppKit.
- CMake 3.20+
- Python 3.8+
- Xcode Command Line Tools
Install with Homebrew:
brew install cmake python3
xcode-select --install- Clone the repository:
git clone https://github.com/J-Liu/PixAI.git
cd PixAI- Build (one command):
./build.shThe script will automatically download and build ExecuTorch xcframeworks from the v0.6.0 release if not present (15-30 min on first run).
Subsequent builds are fast (seconds) — the xcframeworks are cached in Vendor/ExecuTorch/.
- CMake 3.20+
- Python 3.10+
- Xcode Command Line Tools
Install with Homebrew:
brew install cmake python@3.11
xcode-select --installAfter building, you can remove the ExecuTorch build directory (~2GB):
rm -rf .executorch_buildThe xcframeworks in Vendor/ExecuTorch/ are required for future builds.
ExecuTorch xcframeworks (for U2Net watermark detection model):
| File | Description |
|---|---|
executorch.xcframework |
Core ExecuTorch runtime |
backend_coreml.xcframework |
CoreML backend for Apple Silicon |
kernels_optimized.xcframework |
Optimized operators |
threadpool.xcframework |
Thread pool for parallel execution |
Source: ExecuTorch v0.6.0 release (Apache 2.0 License)
Built automatically by ./build.sh on first run.
GNU Affero General Public License v3.0 (AGPL-3.0)
- Real-ESRGAN for super-resolution
- U2Net for salient object detection
Made with ❤️ for my daily use. Hope you like it.