It's not terribly difficult, unless the game is really weird or there are no existing tools to handle it.
Rough steps:
1) Grab a Japanese and English copy of the game
2) Unpack both, this includes "outer" containers like a CIA as well as "inner" containers like a CPK/PSARC if applicable/needed
3) Figure out where the voice files are in both.
4) Replace the English files with the Japanese ones.
5) Repack all the layers you unpacked in [2].
6) Test.
Now, matters are different if there's any of the following problems:
* Proprietary data file formats with no existing tools: you'll need to RE the format to go any further, or hope the game will load files from somewhere outside (many do)
* Weird incompatible structure for voice files between releases. Best example of this is probably God Eater, from what I understand.
* Encryption. While uncommon nowadays, it's possible you'll need a key from the binary.
Essentially, I guess the best advice I can give is to unpack a game, poke around, and read up on the file formats you'll encounter. Audio is often in proprietary formats with odd extensions or something like ADX/HCI which is slightly more standard.