Comment by arecsu

About the uncompressed audio replacements, it makes me wonder how difficult would it be to train a model with a huge (but simple) library of sound effects and samples of high quality, and also feed them their equivalents"low quality" sound signature close or identical to what SNES have. The technical data about the SNES limitations should be there to know how to process these effects as precisely as possible, right? I'm not really a sound guy, so I might be wrong.

Maybe this could result in a much more automated way to re-sample many more sound effects from the SNES massively! Just a thought