We have been looking the other way for years and not at the camera in video calls. NVIDIA just fixed it

  • 42

We spend hours talking on video conferences with people who barely look us in the eye. The webcams of laptops or mobile phones cause almost no one to look at the camera – which is like looking into our eyes – when they speak: how much they talk looking at the screenwhich means that for us they are looking at something up (or down, or to the side).

That is precisely what solution to NVIDIA Maxine, an SDK that allows you to correct the position of the eyes thanks to an artificial intelligence algorithm. With this it is achieved that we are always looking our interlocutor in the eye even if we are looking elsewhere. The idea is fantastic although at the moment it is only available for developers.

The videoconferences of the future are (almost) here

The NVIDIA Maxine SDK has been available for some time now, and offers a number of GPU and AI-powered features that are aimed at improve the quality and characteristics of our videoconferences or live streaming sessions.

The SDK has several sections —audio, video or augmented reality effects— and can be combined with NVIDIA Riva, another SDK that is designed for conversational applications and that, for example, helps in tasks such as transcription or translation in real time.

Just yesterday NVIDIA highlighted several developments in this area at NVIDIA Maxine. One of them was an audio effect that solved the ‘cocktail party‘ what filters out unwanted background soundsallowing us to be heard better.

However, one of the most striking options is the ability to correct our gaze to pretend that we are always looking at the camera and we make “eye contact” at all times with our interlocutor.

In the demo of Alex Qi, responsible for this project, you could see how looking away actually his eyes were always looking at the camera, or so it seemed. The function is part of Augmented Reality SDK from NVIDIA Maxine, which is available to developers in preview.

We are in the era of video conferencing and the webcams of most laptops are a horror

What does that mean? Well, at the moment it is not easy for us to enjoy this option, since it is the developers who will be able to access that function in a restricted way – You must sign a confidentiality agreement.

This solution is currently designed so that companies that, for example, develop video conferencing applications —Zoom, Skype, etc— can experiment with this function and include it in their services in the future. It is of course a great proposal that, yes, will also require an NVIDIA graphics card that gives access to these artificial intelligence functions.

Image | Dylan Ferreira

More information | NVIDIA

We spend hours talking on video conferences with people who barely look us in the eye. The webcams of laptops…

We spend hours talking on video conferences with people who barely look us in the eye. The webcams of laptops…

Leave a Reply

Your email address will not be published.