Skip to content

GazeAwake v0.1.0 — Gaze Awareness for MacBook (Experimental)

Pre-release
Pre-release

Choose a tag to compare

View all tags
@XGXGX1231 XGXGX1231 released this 03 Aug 21:38
· 5 commits to main since this release
v0.1.0
0f3e6ef

GazeAwake v0.1.0 — Experimental Proof of Concept

GazeAwake v0.1.0 is the first public proof-of-concept release of a native Swift MacBook menu bar agent that estimates coarse screen attention and connects it to display power behavior.

Motivation: attention-aware continuity on MacBook

Users who become accustomed to attention-aware display behavior on compatible Face ID-equipped iPhone and iPad devices may miss the same interaction pattern when returning to a MacBook. GazeAwake explores a local, user-controlled approximation: wake an idle-sleeping MacBook display when attention returns, keep it awake while the user continues looking, and restore normal energy-saving behavior when attention leaves.

This motivation does not imply technical equivalence. GazeAwake is independent from Apple, does not use Face ID or TrueDepth hardware, and does not reproduce the complete Attention Aware feature set. It uses a less precise webcam-based heuristic and openly documents its current limitations and resource cost.

Download and install

Download GazeAwake-v0.1.0-macOS-universal.dmg, open it, and drag GazeAwake.app to Applications. The Universal build supports Apple Silicon and Intel Macs.

This experimental DMG is ad-hoc signed and is not Apple-notarized. If macOS blocks the first launch, try opening GazeAwake once, then use System Settings → Privacy & Security → Security → Open Anyway and approve camera access. Only install the DMG from this official GitHub Release.

SHA-256: d00c4fed28b3739d16df44345db9ccd45c62d24f5dce0bc2cec1102477fe9428

Highlights

  • Local AVFoundation + Vision attention heuristic
  • No third-party model, package, Python, OpenCV, analytics, or network service
  • No frame recording, caching by the app, or upload
  • Wake display when confirmed attention returns
  • Keep display awake while confirmed attention remains
  • Restore normal idle behavior when the user looks away
  • Local callback and distributed notification integration
  • Menu bar pause/resume and wake-behavior toggle
  • macOS 13+, Apple Silicon and Intel build support

Experimental limitations

This release is not precise eye tracking. When pupil landmarks are unavailable, it falls back to a face/head-orientation heuristic. It cannot wake a fully sleeping or closed-lid Mac and cannot bypass authentication.

The current landmark mode is resource-intensive. On the documented Apple M5 test machine, active physical footprint was approximately 207 MiB, peak physical footprint approximately 292 MiB, and short CPU samples approximately 21–27%. These measurements are disclosed rather than marketed as lightweight performance.

Next research areas

  • Low-power face-presence mode
  • Sensitivity presets and calibration
  • Launch at login
  • Unit tests and broader hardware benchmarks
  • Screenshots and demo GIF
Assets 3
Loading