Trying to build a visual novel for MSX1 and looking for suggestions.

You can't really load and play a tune. But loading goes quickly, so why do you need that?

Indeed if you use monochrome slides with no full screen, the data to load would be less than 4KB, even less if you compress the bitmap; so, why bother with complex music/trackload routines?.

Anyway, of course it can be done; but you must create your own track loader and make a complex timing scheme to find how much data you can load from disk before you need to feed again the PSG. AFAIK there're no published code capable of doing so...

no. The problem is that any manifacturer used its own disk controller. You shoud do a different code for any possible disk hw. Standard fixed only bios calls

Also... if you want to have music playing while showing the next image, easiest way to do it is to make your game a MegaROM.
Use each 8kb page for a different graphic (or graphics, using some compression) and you'll be able to switch as many graphics you want without having to do some tricky code.

If you reduce your graphics to tiles + tilemap, you can get even better results

If you want to make it look great, I'd go for full colour full screen images. That's 12 KB. And if you want to overlay sprites, that would be roughly 13,440 bytes per image. Adding an efficient compression algorithm could drop that by 50%. Problem here, of course, is that there is no automatic graphic converter that sets up the sprites overlay. If you skip sprite overlay, there are plenty of options out there and they do work great. Say an average of 7 KB/image. In a 360 KB disk, that would made about 50 images. Twice that if you go for 2DD floppies (720 KB).

Once that you have the images, you need some logic. A nice variable width font with a correct blitter routine and the interface that you want to use. And some music and FX would help, of course. Anyway, even if it is an easy project from the technical point of view, it becomes tough when considering the artwork, unless you use the SAC method: Shameless Appropiation of Contents

By the way: dithering is cool if you happen to use an old CRT. If you use a sharp LCD/LED, I'd suggest to go without dithering: plain colours even if there is some bleeding.

Cool initiative! Congratulations for coming with such an idea. I'd like to know more about the plot, if there's already one.