Breaking

Open-Source Project 'face_looker' Uses Generative AI to Make On-Screen Faces Track the Cursor

July 15, 2026 at 05:05 EDT

Open Source
Media Generation
Software Dev & Coding

A newly published open-source project called face_looker uses generative AI to create interactive on-screen faces whose eyes and heads follow a user's cursor in real time. The tool, hosted on GitHub by developer kylan02, pairs an AI image pipeline with lightweight front-end code to turn a single portrait into a responsive UI element.

Open-Source Release · face_looker

AI-Generated Faces That Watch Your Cursor

A new project turns a single front-facing portrait into an interactive UI element — generating a grid of gaze-directed AI portraits, then swapping between them as the mouse moves for smooth, real-time eye tracking.

121

default images per set (an 11×11 gaze grid)

5–8 min

to generate a full 121-image set

input portrait, 512×512, neutral expression

Grid size is tunable — more frames, finer tracking

Image count scales with the step size. Column height = number of generated frames.

7×7 grid
minimum

121

11×11 grid
default

256

16×16 grid
maximum

How it works — one portrait to a responsive face

Input portrait

512×512 neutral front-facing image

→

Batch generation

grid of gaze-directed 256×256 WebP frames

→

Real-time display

frame swaps to follow the cursor position

Lightweight by design

It uses static image switching rather than live inference — keeping the on-page experience fast while the AI work happens up front.

Ships with

React useGazeTracking hook + FaceTracker component, plus a vanilla HTML/JS/CSS bundle and touch support.

Uses

Portfolio headers, interactive avatars and decorative page elements — with preloading tips for performance.

STRENGTHS

Higher-fidelity, smoother tracking from a single image; one-click generation with a free-tier API and drop-in front-end blocks.

OPEN QUESTIONS

Quality depends on the input image, larger grids cost more time, and as a fresh release it has no independent benchmarks yet.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.