MOTUS was a game developed that uses optical illusion effects to render – and this method uses it to display text. Individual frames reveal nothing, but when taken at 30fps, the illusion is revealed.

Would be interested to see if adversarial AI model attacks could be made using this method.