I too was thinking about something like this a few months ago. There were couple of reasons I didn't pursue the idea. One, the image generation AI wasn't reliable enough. Like, I couldn't get it to generate 2 images where the characters looked consistent, let alone a book worth of images. Two, the margins were quite small, so didn't seem like a viable business.

Wondering if you've thought about such things and your perspective.

What do you mean by RTL because all I can come up with is Verilog or VHDL and I'm certain that's not your meaning. I'll try it out. I have a children's book story I've been trying to image generate for 3 years now and it's not yet worked out. I think the primary reason it fails is that the scenery I request is lifelike yet extremely rare to actually see, although, I did see it, and that's what inspires the story.

This is really cool. I wish the example stories let me see the entire book and purchase them if I like them.

I’m skeptical about the stories being good quality so seeing the full stories might mitigate that.