It all starts with source recordings. A number of GT's source recordings are recorded at partial throttle or in neutral, so as to not damage the car they've sourced. That's well and good, but the intakes aren't going to resonate correctly at part throttle, and the engine doesn't pump as much air leaving the exhaust kind of flat as well. This results in that weird mid-rangey, sort of tonal washy sound that GT seems to universally have. Combine that with very few loops that are spread far apart and thus have to be pitch shifted to the point where they sound stretchy and voila, the GT sound. It also would seem that they use very few parameters to drive the engine, for example maybe just RPM and throttle. My guess is that they're memory-limited, and can only fit so many audio files into RAM. They really need to move onto the next gen to see big advancements in their audio.
I wouldn't be at all surprised if the "new sound system" they say they're researching is granular. GT can get the access to the cars they need. But they would only do this drastic of a change with a reduced car count, and it would be expensive and time-consuming to develop this from scratch. If it showed up in GT7 and the car count was 3-400ish, I wouldn't be surprised. They also, maybe, could be researching fully-synthesized, where there's no recordings at all, just wavetables that are dynamically generated. Think of that like a musical synth that happens to sound like car engines, and is infinitely tunable. My guess is that there will be good sounding car games within 10 years that use fully synthetic sounds. Full synth would be extraordinarily cheap to produce a vast array of content for (just need a good reference and one designer to make the appropriate synth patch), but would be a complicated system to develop and integrate and fit into CPU budgets. You could even write some code that does a lot of the authoring work based on physical properties of the engine; cyl count, displacement and stroke length, V angle, and firing order determine the base sound, and then intake length and diameter, header length and diameter, and exhaust length and diameter determine the intake and exhaust resonances respectively.
Back to the comment I quoted, I have no doubt that anyone with a video editor and 10 minutes of free time can take a gameplay capture and a youtube on-board video and overlay great sounding audio. "Sound Mods" for the old simbin GTR2 have been around for a while and people do just this. Making video is easy because it will always happen the exact same way. Game audio is exceptionally difficult because it can only be reactive to the player. You can't just play a youtube video when the car starts moving, because maybe I don't shift when the video shifts, and maybe I want to drift a corner which sounds completely different than if I properly apex and etc. That's where audio designers have found these complicated systems like granular or like loop-based, or hybrids, or synthesized. So no, I almost guarantee you could not just "mix your own sounds and be done with it" and end up with something you'd find enjoyable,
despite what Tiff Needel would have you think.