Audio is no longer optional. Generations now ship with a voiceover script, synthesized and stitched into the video, so the export has a coherent narration out of the box.
In the editor: a voiceover panel per scene. Edit a line, click regenerate, and only that scene's audio gets re-synthesized.
Voices: a curated English and Spanish set to start, picked to pair well with release videos. More languages on the way.