Now that the RTX 3080 has officially launched, the NVIDIA Broadcast app is also now available. Some good news for RTX 2000 series owners; It works for you too! But what does this app do? Is it worth using for streamers? Time to dive in and find out!
Now, a bit of a Disclaimer; I don’t actually own an NVIDIA card to utilize this feature or provide a first-hand experience. All of the information presented here has been sourced from NVIDIA’s official announcement and This video by EposVox. I strongly recommend giving this video a watch.
Table of Contents
RTX Voice was absorbed into the NVIDIA Broadcast App
RTX voice stood at the forefront of AI-powered noise cancellation technologies. Like it’s VST suppression cousins, this was able to remove unwanted background noise from the audio signal. However, unlike those other tools, RTX voice accomplished this task on the hardware level using its tensor cores. Unofficially though, it actually worked on the older GTX series cards as well, utilizing the CUDA core technology. Though, NVIDIA eventually released that to all GTX cards, forgoing the need to do that driver modification.
For those of you that want to try out the AI noise removal capabilities but aren’t ready to upgrade to an RTX GPU yet, we have also patched RTX Voice with support for NVIDIA GeForce GTX GPUs. Though, of course, your mileage may vary on older cards.Nvidia Broadcast app release article
However, the NVIDIA Broadcast app is restricted to RTX only cards. (So far.)
New in this version of the noise Removal is the ability to apply the filter to an audio output device. You’d use this primarily to treat Discord audio, so anyone with particularly noisy backgrounds can be taken care of.
Note: You’ll need to set your discord audio to its own audio channel. Here’s how to do that.
Notable Feature – RTX Greenscreen
If you don’t have the budget for the Lights are a staple of media and content creation. They help create interesting focal points, improve the quality of your webcam or DSLR, and are essential for streamers to ensure consistent lighting. This is especially true for high-action gameplay, where the monitor light can very quickly become an ugly flashing light show if there isn't a key light powerful enough to overpower it. More needed or the room for a green screen, then this may be an option for you. It does have some limitations, in that if you are not well lit, then you will see a slight flicker at the edges of your body. For a point of reference, it looks very similar to what Xsplit’s Vcam is capable of.
However, unlike Vcam, this effect does not come with that large CPU performance hit. Instead, the program simply offloads the processing onto the Tensor cores on the GPU.
…unlike Vcam, this effect does not come with that large CPU performance hit. Instead, the program simply offloads the processing onto the Tensor cores on the GPU.
Notable Feature – Background blur
Expanding on the Green screen effect is a background blur effect. This creates a faux bokeh depth of field effect, Which looks a little like this:
This is an example image from another article, The image shows a very blurred background effect, which puts the foreground elements into hyper-focus. However, like the green screen effect, it is subject to edge artifacts all the same. Supposedly, if you already have a weak Depth of field effect, those artifacts are blended in a bit and are less noticeable. Thanks to EposVox for pointing that one out.
Notable Feature – Autoframe
This feature is particuluarly handy for VR creators who are moving around a lot. With this feature, your camera will track your face. No word on its ability to track with a HMD on it just yet, but I imagine it would be able to do so.
Unfortunately, you are unable to use the Autoframe feature alongside RTX Greenscreen & background blur. Its one or nothing. Perhaps in the future, in a later release? Time will tell.
Provide Samples to improve the NVIDIA Broadcast App Features
A few of these features are marked as “Beta” releases. Like RTX voice before it, the effectiveness of these features is likely to vary wildly. If you want to help the software improve, you may provide samples to them for the purpose of training the AI.
Take those tin-foil hats off. They aren’t spying on you. You provide these samples willingly, nothing is collected against your will, and they will only be used to improve the software!