Ghidra is a high end hacking tool
https://ghidra-sre.org/
Indeed it is the NSA's answer to the likes of
https://hex-rays.com/ida-pro/ (the main industry program) and
https://rada.re/n/ (the other big open source offering).
I don't tend to use it so I am not entirely sure what abilities it will inherently have to parse this format; as file formats are so commonly custom then all those programs will feature options to define custom ones, and aid in the documentation of formats. It is curious to see full directory names like that though as most formats will not include such a thing for every file and instead such a thing gets created if necessary (massive waste of space and effort to scan it if you store it all like that for no great gain) but also not that hard to generate by the would be hacker.
http://wiki.xentax.com/index.php/Game_File_Format_Central https://wiki.multimedia.cx/index.php?title=Category:Game_Formats being examples of formats found in games if you did want to see how such things have been done in the past; there are certain types of data you will need when defining an archive format so most of the time figuring such a thing out is figuring out the order and data types used for those particular things and whatever other features came along for the ride.
I cover pulling apart some things in
https://gbatemp.net/threads/gbatemp-rom-hacking-documentation-project-new-2016-edition-out.73394/ , though you might find you encounter sector sizing in this (32 bits does do 4 gigs easily enough though so you might be lucky) rather than more direct methods common in older/smaller file size efforts. File names, file sizes, file locations, subdirectories, compressed status... all being things to ponder in this (file size and file location can be mutually exclusive as you can use one to calculate the other, though sometimes both are good if the file locations are not one after the other and there is some padding in the middle).
16000 odd sound files... possible I guess -- if there are say 20 player characters and they all have a variety of individual attack sounds, hit sounds, low health... and then all those got doubled up to sound more fitting in a sewer level, possibly plus a few languages then it can end up there, and PS1/PS2 having the storage it did it is not like cartridges where space was at a premium.
"take them as they are". Does that mean you can only extract them as .wav or something rather than raw data you could search the file for? Pity if so as that can really help in searching for things. Also yeah sounds like you will need to either have a format description for this particular thing called img, though
Re hex editor for pulling files out. If you ultimately know enough to use in a hex editor you probably also know enough to get
https://web.archive.org/web/20170218180937/http://min.midco.net/cracker/filecutter.zip to do it for you in a large batch file. Won't help so much in putting it back together though, however if it is going to be a simple pointer/file location and you can overwrite things then you might not have to build a full file builder tool or do a lot of manual work. You might even get lucky and find the Japanese or whatever are smaller than the things you are replacing in some/most instances which further reduces the effort. One minor concern is if the Japanese is higher quality than your target (happens from time to time, especially in European versions aiming to squeeze in say 5 languages) in which case you have other considerations (drop quality, make game understand higher quality, scrap the whole effort and port the game's text files to Japanese/whatever version instead).