Nvidia-backed AI Makes Realistic Avatars for Businesses

An Nvidia-backed company, Synthesia, launches AI avatars for business video production.

The “Expressive Avatars” introduced by Synthesia, a video creation startup fueled by artificial intelligence, simulate human-like emotions in response to user input. The Nvidia-backed company states that enterprise users who wish to create on-demand training videos, presentations, and other corporate materials without assembling human actors will have access to the avatars. 

Thursday via LinkedIn live-streamed a virtual event at which the CEO of Synthesia, Victor Riparbelli, displayed the avatars Alex, Joshua, Talia, Francesca, Jaz, Julia, and Paloma. Seven digital actors imitate humans of diverse ethnicities, ages, and accents, as well as dress, with some adorned in business attire and others in clothing suitable for the workplace. How the avatars deliver their scripts is customizable by the users, including the location of their glances, the duration and nature of their pauses, and the tone of their expressions. In addition to providing scripts in 135 languages and a male or female vocal, the avatars can lip sync to audio clips uploaded by users. 

Additionally, users can modify the video’s backdrop. Synthesia provides examples such as the interior of a retail establishment, a bustling office environment, and two-dimensional visual assets. However, one could also envision the avatars being implemented in simulated automobile dealerships or on airplanes to film travel safety videos. Users can produce photorealistic, closed caption-optional videos in minutes without investing the time, money, and effort required for traditional video production, including cameras, microphones, human actors, and laborious edits.

Still, the avatars are intended for something other than casual users. In light of the heightened global scrutiny of misinformation, Synthesia wants to limit the accessibility of its avatars exclusively to enterprise clients. Although standard web users can preview the avatars for a maximum of 36 minutes of video per year, accessing Synthesia’s complete collection of 140 avatars, collaboration functionalities, custom typefaces, and additional design tools necessitates a paid account. Enterprise access also allows users to generate a “digital twin,” an avatar resembling the user. Synthesia hopes that by primarily targeting large businesses and implementing a paywall for most of its customization features, it can effectively mitigate several misuses commonly associated with deepfakes and other “virtual characters.” 

More than 55,000 organizations purchase technology from Synthesia, primarily to produce corporate presentations and training videos. The company secured a funding round of $90 million last year, led by NVDA.O NVentures, which Nvidia owns. Synthesia is merely one of the recent AI-related peripheral investments made by Nvidia. The semiconductor manufacturer acquired the startup Run:ai on Wednesday.