Nvidia improves the sound of webcam microphones using AI

[10:14 Tue,11.February 2025 by Thomas Richter]

Nvidia has released the latest version of its Broadcast app, featuring two new functions – Studio Voice and Virtual Key Light – that enhance both the audio from microphones and the video from webcams via virtual lighting. The update also includes optimizations for existing effects and a redesigned user interface.

The great thing is that the GPU-accelerated AI features of the Broadcast app can be used in all programs that use a webcam and microphone for input, such as OBS Studio, Streamlabs, XSplit Broadcaster, Twitch Studio, Skype, Zoom, Microsoft Teams, and Google Meet. Nvidia Broadcast is simply integrated as a virtual device.

studio-ai-geforce-rtx-50-series-broadcast

Studio Voice and Virtual Key Light

The new Studio Voice function improves the sound of a simple microphone, such as that of a webcam – according to Nvidia, it should sound like a high-quality studio microphone.

The Virtual Key Light function provides even, post-processing lighting of the recorded person, as if a real key light, a main light directed at the person at an angle, were used to highlight the shape and depth of the image.

The Verge has tested both new features.

Nvidia Broadcast Studio Voice

Further Improvements

The new version of the Broadcast app also improves some of the previously offered AI functions, such as voice quality via background noise suppression, and ensures a more natural appearance in the Eye Contact function through small, random eye movements. Foreground and background separation is also improved when using virtual backgrounds.

The user interface of the NVIDIA Broadcast app has been fundamentally redesigned, now allowing more effects to be applied simultaneously and offering a side-by-side camera preview, a GPU utilization display, and more.

Nvidia Broadcast Virtual Devices

The AI Features of the Broadcast App

This means that the range of AI effects possible for live webcam images with Broadcast app 2.0.0 now includes suppression of microphone noise and room echo, improvement of microphone sound, virtual backgrounds, suppression of video noise, post-processing lighting via key light, webcam auto-framing, and post-processing modification of the direction of the eyes (Eye Contact function), so that they seemingly always look directly at the viewer. Nvidia aims to make its high-end graphics cards appealing for applications beyond gaming, video editing, or AI for users by utilizing their enormous computing power to upgrade simple webcam setups.

Both new effects are still in beta and require at least an Nvidia GeForce RTX 4080 or 5080 GPU or higher. They should not be run simultaneously with CPU-intensive programs such as games.

more infos at bei blogs.nvidia.com

deutsche Version dieser Seite: Nvidia verbessert den Ton von Webcam-Mikros per KI