RTX Voice: Goodbye, Background Noise

RTX Voice

On the day of this writing, I woke up to a Twitter post shared on a group chat. It was a tweet by theGunrun, a former employee of Twitch. It was a demonstration of the new RTX Voice, a piece of software by NVIDIA that uses AI technology to process input audio on-the-fly with your GPU, removing background noise from one’s voice. Users are reporting amazing results from this new application, and I can attest to its effectiveness.

If you’ve done voice chat with enough people on Skype, Discord, Facebook Messenger, Viber, or so on, you’d know that most people use either cheap microphones and headsets or just the built-in microphone on their webcams. Even people who just use the microphone on their smartphones will give you an idea on how much background noise can affect sound quality.

There are solutions like noise gates in applications like OBS or Voicemeeter (which I use), but most people may find that to be too complicated for them. RTX Voice is an easy-to-use solution, but its effectiveness is such that even people who know what they’re doing may find it useful as well.

What is RTX Voice?

RTX Voice is the latest among many things that are showing how NVIDIA’s focus on developing AI technology is paying dividends. They already have DLSS—Deep Learning Super Sampling—which is doing a lot of amazing things for game performance.

This application makes use of the GPU of an NVIDIA graphics card to process sound picked up by your computer microphone, distinguishing between ambient noise and one’s voice. The ambient noise is then made silent while human speech is kept, thus improving quality of communicative audio.

NVIDIA GPUs and software already do amazing stuff like NVENC encoding and Shadowplay gameplay recording, both of which have been boons for streamers and content creators who need a rendering solution that produces little to no CPU overhead.

As my friend BJ David of RoninVampire and Bistek.ph stated in a Facebook post:

NVIDIA’s improvements in DLSS and RTX Voice really highlight the difference in technological approach between them and AMD.

If the task was to help you climb a tree, NVIDIA would make you a mobile auto-escalating ladder that auto-adjusts its height and controls its speed depending on your weight. On the other hand, AMD will build you a cannon that will shoot you into the stratosphere and hope you land on the treetop.

Here he is, trying out the application for himself:

RTX Voice adds to that growing suite of hardware and software solutions for game-centric content creators, especially in the current global COVID-19 quarantine lockdown.

Why I Need RTX Voice

At the moment, I’m building my own recording booth for voice-over in an empty closet. However, the emptying of that closet has rendered the space in my home gym almost entirely unusable for its intended purpose due to all the decades-old detritus I had to remove and will remove, especially without help during the current quarantine lockdown.

But the recording booth is crucial for getting more voice-over work, especially as a freelancer. Most clients forbid post-processing and require the highest possible quality for voice recordings, so I have to invest time, money, and effort in building my own booth with proper acoustic treatment. I even bought foam from China just as the COVID-19 scare was building up.

Background noise is the bane of anyone with a condenser microphone. Due to the sensitivity of the hardware, you can even make it pick up the tic-toc of a wall clock if you crank the gain high enough. That’s the trade-off for having high-quality audio recording, and it’s a nightmare for singers and voice-over artists everywhere.

It’s usually remedied with acoustic treatment and doing your recording in a quiet room, but not everyone has that luxury. If you live in a small apartment in the city, then you’re pretty much fucked. You can try hardware and software solutions like a noise gate or noise removal in post, but it does janky things to your recording that make them not worth the effort.

RTX Voice for GTX (Non-RTX) Cards

Despite the name, there’s actually a way to use RTX Voice with a GTX 1600, 1000, or even 900 series card. It does require some finangling of installation files, as detailed in this guide on the Guru3D forums.

My current graphics card is an NVIDIA GTX 970, and it seems like RTX Voice works pretty well. Perhaps it doesn’t work as well as with a more advanced card as it doesn’t totally eliminate tapping sounds and other loud rhythmic noises, but it does well enough in removing humming noises like the fan, the air conditioner, and even traffic outside.

I’ve always had a problem with outside noises as I live near a busy road that’s notorious for traffic jams. Once this quarantine lockdown is lifted, things are going to get crazy again. If RTX Voice really is as good as advertised, it will become a much less significant problem for me.


There’s a lot of criticism regarding NVIDIA’s business practices, but you can’t fault their R&D. While AMD cards are still better in terms of raw price-to-performance, NVIDIA cards continue to be the better option in terms of features and efficiency. Developments like this convince me that my next upgrade will be with Team Green again (likely an RTX 2070 Super).

Yeah, I’m a bit of a fanboy. Bite me, they make good shit. If AMD starts making better graphics cards, I’ll go Team Red. I plan to get a Ryzen 7 anyway.

Meanwhile, time will tell as more people try out RTX Voice if it’s not just for streamers and frequent users of VoIP software, but for serious musicians, podcasters, and content creators as well.

Got Feedback?

Have something to say? Do you agree or am I off-base? Did I miss a crucial detail or get something wrong? Please leave whatever reactions, questions, or suggestions you may have on the comment section below.

You may also like/follow and leave a message on either Facebook or Twitter. Please subscribe to both the Avoider.net YouTube channel and my personal YouTube channel as well for more content. Thank you for dropping by.