Is using voiceover adding any value to e-learning?

The research supporting audio is discussed in eLearning and the Science of Instruction. Clark and Mayer call this the "modality principle," and it essentially says that people remember better from words spoken as narration than as on screen text, especially if you're using accompanying graphics or animation. In one study cited, students who heard narration produced twice as many solutions on a subsequent test as students who read identical on-screen text. Another study showed a 29% increase. The modality effect matters more for less experienced learners--if you're training doctors or engineers, it might not make as much of a difference as if you're training retail or hospitality staff. It makes more of a difference for complex material at a rapid pace than easy material at a slow pace (or where learners frequently pause or control the pace).

You are correct, however, that you should not have the voiceover read the on screen text verbatim. You might have the graphics and a few key phrases or maybe one sentence on screen, and the rest of the text is in the narration. Having the voiceover duplicate on screen text is called the "redundancy effect" in the research. You should either have on screen text or audio, not both. Multiple studies have shown that redundancy leads to lower transfer and application of problem solving.

If you search for the specific principles above or for "multimedia learning theory" or "multimedia learning principles," you'll find plenty of articles on the topic. You can also search for "dual coding theory," which is the underlying psychological theory. Research on "cognitive load" will also sometimes be relevant, although a fair amount of that is about other related topics.

You can start here with Google Scholar. There are over 1 million hits. https://scholar.google.com/scholar?q=multimedia+learning+theory&hl=en&as_sdt=0&as_vis=1&oi=scholart