In fact, there’s a whole SIGGRAPH video detailing this procedural tech with CD Projekt RED’s Mateusz Pop?awski explicitly stating that the team wanted to create better lip syncing than the Witcher 3 with “muscle-driven emotional expressions.”
As the video shows, it genuinely does work. The tech takes into account brow, neck, mouth, eyes, and emotional movements into consideration to produce the final animation. There’s even a clip of this working in-game starting at 14:51 in the video below.
The tech also works in 10 languages for Cyberpunk 2077 using custom pronouncing dictionaries for all languages, and grapheme-to-phoneme (G2P) models for all languages. This allows the prediction of out-of-vocabulary words. You can watch the full video below: