Created using the first known Deepfake generator in 2003, drawing from a library of phonetic samples and visual expressions spliced to form new words with corresponding mouth shapes.