Dialogue bubbles and narration

My go-to answer is the ever unhelpful "it depends." Everything you describe can be its own optimal design choice in at least a few different projects, with particular scopes and deadlines and capabilities. I see this kind of content as living on a continuum, from super static on the left (perhaps a wall of text in a slideshow with a single portrait of a person) to fully dynamic (a video of characters emoting narrative performances.)

Speech bubbles are often fine, and most people are already conditioned to understand their meaning, so that can make development easier. Instead of reinventing every wheel, incorporate what people already know.

Typically a disembodied narrator passively describing what's happening in a scene is considered even worse, so stay clear of that where you can't go without some kind of audio engagement. It's so hard to do well in narrative structures that it's considered a major red flag when reviewing film script submission.

I wonder if there aren't more criteria you're working with, or perhaps some other part of this that could be framed as a problem or obstacle? The method you describe can succeed purely on style and art, so I'm curious where you see its gaps.