Currently the voiceover mostly read the text on the screen. In a really polished video however, the voiceover complements the visuals.

I was able to achieve this by using a separate llm to give me voiceover snippets that do this. But it was quite time consuming. It would be 10x better if I could directly in Supademo give the voiceover AI instructions on what kind of content it should make. (e.g. complementing the visuals)