I wouldn’t say load instantly: on a fast wifi it has similar times to what you describe, the first api request for data starts after 4s and then you have the requests for the images. Now almost all the images are cached and served by a service worker (not sure if this is standard behaviour of weweb or if it has been added) so when you load the page the second time the images are loaded instantly after the first delay for the initial api request. And the average image size is less than 20kb. Webp helps a lot and is probably the only other image optimization needed for you.
Instant is more like something like https://movie-app-with-remix.vercel.app/ (the first image starts loading at 0.5s) but this is with code and server rendered, so it’s a completely different tech.
try with webp when you optimize the images and see if the result is better