Text heavy slide creation tools (such as LaTeX and Markdown based ones) are aimed to make slide creation easier, not actually watching the resulting presentation. It's not to say that you cannot make nice slides with those tools, but using systems that make it much easier to write text than to visualize really promotes the wrong thing for presentations.
People don't tend to notice this as much though, and they are more concerned with familiarity or speed-of-production of the slides rather than make the slides effective for the audience. But hey, that same issue can be raised for most PowerPoint presentations, too. People just don't know how to make good slides and just take whatever is easiest for them.
This does allow you to embed images and diagrams. I think they rather mean the slide description language is text-only.
But even embedded images implies uninspiring slides to me.
For example they have a system that allows you to add source code to a slide, but then don't you want to be annotating that code? Adding arrows, lines, highlighting bits, breaking it apart, etc?
For a recent presentation I also tried reveal.js again and I really liked it overall (compared to LO Impress). I do science, so I'm not that into annotating source code, but this total lack of presentation "spice" also really bugged me and I basically ended up writing html with semitransparent overlays over key graphics (more of an explorative presentation)...
As to this: I just don't get what's so special about this. I especially don't get, what this could do, what plainly putting your presentation into git doesn't do already? Apparently behind the curtains (which probably does not concern people who do `curl <url> | bash` stuff...) it's using pandoc to convert md to a reveal-js presentation and using a node-based live-server for display. Not even sure if speaker notes work (or if you have to use the reveal.js internal server for that)...
Also, this is a security nightmare, as a ton of boilerplate-dockercode/npm-packages/random-binaries is necessary for just about nothing, as in the end it's just downloading the pandoc-binary from github to a docker-container (well, building pandoc was to hard apparently). I'm scrambling my head, what this means for security in applications developed by people lead by the poster if he goes forward like this.
This is something that I've been intrigued about too. Most of the cli/text based creation tools out there focus on text based content as well. But, for many people myself included, the desire is really for a text based creation tool that focuses on image or diagram based output. Features like auto tiling (with priority) of images on a slide would be useful.
So many people end up writing what should be their presentation notes as their slide content. Personally I think you're better off with a holding image and move your content to the presenter notes section.
I've used revealjs.com and hovercraft [1] in the past but usually end up on GSuite Slides to throw things together quickly. Slides (like PowerPoint) can't do branching presentation paths though :(
People don't tend to notice this as much though, and they are more concerned with familiarity or speed-of-production of the slides rather than make the slides effective for the audience. But hey, that same issue can be raised for most PowerPoint presentations, too. People just don't know how to make good slides and just take whatever is easiest for them.