Ventriloquist Effect: Fusing Sight and Sound by Precision Weighting (2D)
The same Bayesian cue-combination model on a canvas: a puppet and a loudspeaker on a stage, the likelihood of each sense, the fused posterior, simulated trial percepts and the predicted scatter, with the numerical checks of the shared model.
Why does the voice of a dummy seem to come from its lips? Place a puppet and a loudspeaker on a stage, set how blurry the eyes and the ears are, and watch the brain fuse the two position cues by precision weighting. See the vision and hearing weights, the mean percept pulled toward the sharper sense, the trial-to-trial scatter smaller than either cue alone, and a histogram of simulated trials. A Blender-built stage, dummy and speaker, with numerical checks against brute-force grids and Monte Carlo.
The same Bayesian cue-combination model on a canvas: a puppet and a loudspeaker on a stage, the likelihood of each sense, the fused posterior, simulated trial percepts and the predicted scatter, with the numerical checks of the shared model.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install