Dubbed RTX Voice, the software is developed for Windows 10 and uses processing power from Nvidia RTX GPU Tensor cores via CUDA programming API to run complex machine learning and artificial intelligence algorithms.

This AI stuff works out what is a human voice and what should not be there and then deletes the rubbish.  We ran it on a Trump (inject disinfectant in your vains) press conference and it was completely silent other than the reporter's questions.  Actually we made that up, but it is really useful if you are trying to record something in a noisy environment.

This means you can live stream or record videos in noisy environment with things like phones, vacuums, lawn mowers, leaf blowers, fans, farts, air conditioners, mechanical keyboards, dogs barking and banging on your desk, etc all being silenced from your viewers with only your voice getting.

You can see it being used in anger, below.