Traditionally, powerful GPUs are associated with visual rendering technology, either for playing games with rich high fidelity visuals, or for photo and video processing, or other image-related tasks. But with this new piece of software, Nvidia is pushing for new ways to use their high-end GPUs to power a new audio processing technology that could change the way we use our computers to record audio.
Nvidia RTX Voice is a new GPU powered audio real-time audio processing technology designed to remove background noise from voice recordings, and it seems to be extremely effective. We’re getting into “sufficiently advanced technology is indistinguishable from magic” territory here.
This new technology is still in beta, but as you can see in this video by streamer Barnacules Nerdgasm, it’s able to successfully filter out noise from a loud fan and even hammering on the table.
Warning: This video does include some profanity.
To my ear, it doesn’t sound absolutely perfect, there’s perhaps a slight flat sound to the voice that is being introduced by this technology. But it’s clearly doing its job as intended, and the loud background noises are nearly completely removed from the recording. It’s still just in beta, so it’s entirely possible that this could get better and better as Nvidia work to improve it.
As is pointed out in the video, there are perhaps things that a professional audio engineer could do to clean up audio recordings given enough time, but that this being done in real-time without any kind of human input is extremely impressive.
There’s a whole host of possible applications for this kind of technology, improving the audio quality of commentary for streaming video, improving the clarity of in-game voice chat, making video and audio calls clearer.
Currently, supported applications are:
Each different application has a slightly different setup procedure, for example, Nvidia recommends that you disable Steam Chat’s built-in noise-canceling feature for the best experience. Details for how to use this beta software are posted on Nvidia’s website here, and Nvidia says that you’ll need to update to version 410.18 or above of their drivers, you’ll need to be on Windows 10, and intriguingly you need to have a “GeForce or Quadro RTX” GPU.
Intriguing because the folks over at Tom’s Hardware discovered that you do in fact not need to have an RTX card. They’ve found a workaround to get this software working on GTX 10 series cards. It’s a pretty straightforward hack, and it’s simply removing a line of code that prevents the software from running on “unsupported” cards. Kind of a cheeky move from Nvidia, perhaps using technology like this to drive GPU upgrades, rather granting access to it for all owners of any cards capable of running it. Or perhaps they’ve just not gone to the effort of testing it on earlier cards, and there might be performance issues that develop over time not immediately obvious.
If you do have one of the unsupported GTX 10 series Nvidia cards, you may want to install the beta version today, since there’s no telling whether future versions will still run on unsupported hardware, or whether Nvidia may implement a more rigorous technique of stopping it from running on older cards.
Is this something that you might want to use? Would you consider upgrading to an RTX card specifically for this type of functionality? Are there any streamers that you enjoy but get frustrated with background noise that you’ll be recommending this to?