Voiceover turns the words in a Pattern into spoken audio and lays it under your video.
Local by design
The speech comes from a model running on your own computer. Your script is not sent anywhere. knitted talks to it through a standard speech endpoint, so a base address is all it needs.
Mixing
The voice sits on top of any music, and the music dips when the voice speaks. The final loudness is matched so your reel plays at a sensible volume.