Presumably, they did not record each sentence respectively but have the audios for some sentences compiled from various snippets of these words. I mean, they did this voluntarily, and recording each sentence separately would have taken a whole lot of time not everyone of them is ready to invest into a mere hobby.
Actually for most courses no one records anything, it's a text-to-speech synthesizer audio that is generated automatically. Only if no such audio is available for a given language, then real people record it (from your languages, that's surely Hawaiian and my guess is maybe Irish).
And the TTS technology... it's very far from perfect, I'm afraid. Especially in terms of intonation.