Play the audio of ba over video of a mouth saying ga and most people hear da, a sound present in neither channel. Knowing the trick does not switch it off. McGurk and MacDonald published the effect in 1976 and it remains the classic demonstration that the senses merge into one percept instead of running side by side.
The effect is perception doing sensor fusion below awareness, weighting the eye’s evidence about speech against the ear’s. Dubbed films and laggy video calls feel wrong for exactly this reason.
For anyone building synchronized audio and visuals, McGurk is the reminder that timing is not cosmetic: the audience’s brain is fusing the channels, and misalignment changes what they perceive.