That's assuming you need all three times of day in every location though, which is only the case for her room (and the lighting there doesn't actually change, so that would still only be one set of renders). It would be a few more images, sure, but it would look so much better. Especially for Naomi, who presently has serious problems if you trigger one of these minor scenes in the morning – sitting in her room and working out in the attic both have different outfits, and the sprites are way too brightly lit for the attic.
If we look at Naomi, adding her three other outfits for the layered sprite method, that's 6 emotions × 4 outfits compared to 6 emotions × 6 outfit-location-time combinations. That isn't all that many more, and with a library of these renders to call on some story scenes could be done with these and just a few unique renders which would lead to savings in the future. It'll be interesting to see how the animated sprites go, but I wouldn't dismiss generic location images completely. I mean, you're already doing this in some places, right? It would just be an expansion of that.