I submitted my 180,000 word book to book baby having inserted images with the captions embedded alongside the pictures within the images. My first submission wound up being a 13.5 meg docx file. I rejected the proofs because the embedded caption text had bad edges and was not readable. And the Epub file was 8.03meg which I considered too big.
I reworked the images using different proportions (1500 x 1650px) and a larger font. And reduced the average file size from 2meg each for the first submission to 800k for the second. A total reduction form 58meg of pictures to 12meg. Unbelievably my .docx file size went up to 18.5meg and my new Epub file is 9.7meg. There were four more images inserted for the second go due to the reworking of proportions but