I Tested Nvidia’s AI Tool That Will Make Your Webcam Better, and Wow
When Nvidia introduced the free Nvidia Broadcast app in 2020, it promised to use the AI capabilities of RTX GPUs to enhance the video and audio quality of users’ recordings, provided they had the right hardware. This can be useful when a laptop’s built-in webcam and microphones aren’t able to capture the prettiest footage or richest audio, and could potentially help streamers get by with a cheaper setup. With version 2.0 of the Nvidia Broadcast app, released in late January , the possibilities are even greater… maybe even a little too close to the sun.
In Nvidia Broadcast 2.0, microphones still have options to remove room noise and echo, but the beta now also includes a studio voice effect that “enhances the quality of your microphone to simulate a high-end recording studio.” On the video side, the app can still make changes to your background (replace, blur, or essentially green screen it), remove noise from grainy frames, zoom in and automatically keep you in the frame and, oddly enough, make it look like your eyes are looking into the camera. But the update includes virtual lighting that will help make you stand out better in your videos.
Some of these new features require powerful GPUs. Studio voice, video noise removal, and virtual key lighting all require a “high-performance GPU” and are not recommended for use during gaming or other GPU-intensive applications. All features require RTX hardware, meaning you’ll need at least an RTX 2060 or higher to try them out, but for features that require a “high-end GPU,” Nvidia says it “requires an RTX 4080, 5080 or higher.” However, I was able to run both features on the RTX 4060 mobile GPU.
Nvidia Broadcast in action
The idea behind these AI features is cool, but how well they actually work is questionable. First, they really can be as demanding as Nvidia says. Running the virtual key backlighting feature or the studio voice feature on my RTX 4060 laptop showed that the GPU was completely destroyed by the process. Nvidia’s built-in GPU usage monitor was showing red, with the RTX 4060 near max, and the performance overlay showing it drawing 60W. My laptop’s fans even got louder, as if I was gaming at full power. So from an economic perspective, these features are expensive no matter how you look at them. To run them you will need powerful hardware and then hard run that hardware. Plan to use these features on a desktop computer or with a connected laptop.
But there’s an even more important question: what do they actually look and sound like? Let’s start with the video.
The Eye Contact tool, although available before Broadcast 2.0, is now out of beta. But I’m not sure that’s how it was supposed to be. Of course, turning this feature on makes it look like I’m looking into a video camera. But during my testing, it consistently gave me big blue eyes that made me look like a White Walker straight out of Game of Thrones . For the record, I don’t have blue eyes. Even when I looked at the camera, Nvidia Broadcast still insisted on editing my eyes to be blue.
The virtual key light did what it said. He created artificial lighting to increase the brightness for me without increasing the brightness of the entire video. However, I was not impressed with the results. When it’s on, I just look like I’ve become radioactive. The lighting is very unnatural.
As for the sound, at first glance it sounds quite impressive. The microphones on my laptop are not very good. Even in a quiet room they produce a sound that seems distant and slightly muffled to me. If I turn on the studio voice, my voice sounds much fuller and clearer. But if you listen closely, you can see a strange digitization. It’s hard to characterize, but it doesn’t sound like it’s really my voice. It’s more like the recording of my voice was used in a voice cloner and then repeated everything I said. It’s all a little unnatural and shaky. Listen below:
The studio voice feature also cannot save the microphone from poor recording conditions. When tested in a small room with a fan running at full blast, audio clarity was significantly improved over raw recordings from the microphones, but it was still audibly processed, and attempts to suppress background noise made my voice sound particularly strange.
If you have a decent microphone, studio voice can even make things worse. I made additional test recordings using the Audeze Maxwell headset’s built-in microphone connected directly to my laptop. In both a quiet and loud room, it provided a loud, clear and full recording of my voice without the inclusion of the studio voice. In both cases, the inclusion of studio voice resulted in subtle digitization that not only made the audio sound worse, but also made it more difficult to understand.
Can Nvidia AI replace proper streaming setup?
Given the hardware requirements, performance requirements, and quality of results, the stars really need to come together to make these new Nvidia Broadcast features feel truly useful. If you have an Nvidia based system, be sure to play with this tool. Some features may come in handy, such as automatic cropping. But I wouldn’t recommend shelling out for a new Nvidia GPU just to save on audio and video recording hardware, especially if you want to get closer to professional quality. And don’t forget that the power consumption of the GPU trying to run these functions will increase over time.
The audio quality I got from studio voice over (perhaps limited by the RTX 4060 in my system) wasn’t something I wanted to share with any audience on a regular basis, and it paled in comparison to the quality I could get by simply having a headset with a boom mic. I’ve tested a lot of gaming headsets, and even very cheap wired headsets with a boom mic are much better than what I’ve heard from studio voices.
The eye contact feature has caused nothing but anxiety, and I don’t think it will fool anyone into believing you’re actually looking them in the eye. And a virtual key light, like a studio voice, is not a good replacement for a real key light, especially when available LED lights are a dime a dozen.