Synthetic Faces with Cross

Analyze the Flashed Face Distortion Effect and peripheral adaptation mechanics.

DEMO MODE ACTIVE Touch Anywhere to Terminate Sandbox Demo

Simulation Engine

Align Eyes (Illusion Boost) ON Fix retinal eye coordinates

Sound Output OFF Enable Auditory Cues

Morph Interval: 120 ms

Base Face Size: 100 px

Asymmetry Bias: 0.15

Overview

The Synthetic Faces with Cross tool explores the cognitive limits of visual processing and structural contrast adaptation. Specifically, this simulation models the foundations of the Flashed Face Distortion Effect (first described by Tangen, Murphy, and Thompson in 2011). When individuals stare directly at a central fixation cross while highly varied human faces are flashed rapidly in their peripheral vision, the peripheral faces appear Grotesquely distorted, exaggerated, and deformed.

Because the original illusion relies on the deep, highly specific visual pathways our brains use to parse human faces (such as the fusiform face area), simplified cartoon or abstract vector designs can sometimes dampen the effect. This occurs because highly stylized geometries do not trigger our face-processing expectations with the same intensity as photorealistic human portraits. By introducing strict spatial constraints and toggleable features, this simulator demonstrates how specific structural parameters affect the perception of abstract representations.

How to Use

To maximize the chances of observing the distortion effect using synthetic designs, follow these configuration guidelines:

Stare at the Center: Keep your gaze fixed strictly on the white crosshair in the center of the viewport. Avoid looking directly at the faces on the left and right sides.

Toggle "Align Eyes" (Illusion Boost): When enabled, this option forces the eye locations of the generated faces to remain perfectly stable across frames. Locking the retinal coordinates of the eyes allows the adaptation distortion to target the exact same neural receptive fields, dramatically strengthening the visual adaptation effect.

Adjust the Morph Interval: Keep the transition speed relatively fast (between 80ms and 150ms) to ensure continuous retinal adaptation without allowing your peripheral vision to resolve individual details.

Vary the Asymmetry Bias: Increase this parameter to introduce structural differences between individual facial halves, enhancing the perceived eccentricity of the generated faces.

Technical Details

This processing engine implements several design measures to balance mathematical performance with perceptual accuracy:

Coordinate-Locked Eye Calibration: When "Align Eyes" is checked, variables controlling lateral spacing (eyeOffsetX) and height (eyeOffsetY) are decoupled from global randomness and locked based on facial scale. This eliminates displacement noise, allowing the brain to contrast facial features on a stable baseline.

Enhanced Geometry Shading: Radial visual gradients and high-contrast lines are drawn on an HTML5 canvas using native rendering functions. These gradients mimic basic natural lighting, stimulating depth processing and visual pathways more effectively than flat shapes.

Low Latency Synthesizer: Integrates a Web Audio API-driven tone generation system that tracks mouth and pupil structures, rendering quick acoustic feedback while ensuring the browser main thread remains unblocked.

Future Directions

Future revisions of this scientific simulator plan to integrate high-fidelity generative facial models, allowing users to toggle between vector-rendered cartoons and photographic textures. We also plan to integrate an adjustable eccentricity slider to change the lateral distance of the face centers from the fixation cross, helping researchers map the exact field-of-view thresholds where distortion effects are most pronounced.