If you haven't heard of RTX Voice, it's basically "Krisp" in Discord but with less shitty audio quality. It's based on Tensor, which can fall back onto CUDA hardware, but runs best on RTX due to the hardware accelerated Tensor cores on it. You can find a guide on how to set it up here on Nvidias own website.
I'm currently trying to integrate it as a VST, though progress on this may be slow due to actual work, and StreamFX having priority. But once I have a working version it will be available for free for any Patreon, Github and Twitch supporter - or available for a one-time fee of 60 EUR.
I'll make another post once it is ready.