An artificial intelligence (AI) start-up has unveiled a series of lifelike avatars that can convey human emotions via a user’s text prompts.
London-based firm Synthesia showed off its “expressive avatars” in a video uploaded on the social media platform X on Thursday, showcasing a range of emotions from happy, and sad to frustrated.
The technology comes at a time when deepfake videos raise concerns about avatars convincingly impersonating anyone for scams or misinformation.
However, Synthesia told the MIT Technology Review that the company prefers the term “synthetic media” instead of deepfake, which denotes negative affiliations.
The company, which is backed by Nvidia, says it aims to cut the costs of video production by eliminating the need for actors, long edits and cameras, among other costs.
The avatars are trained on the movements of real-life actors who read scripts in front of a green screen. Amazon Web Services used the technology on Wednesday to clone an avatar version of AWS’ managing director Tanuja Randery, which was shown off at an event.
Synthesia’s technology is used by 55,000 businesses to make digital avatars for such company presentations, the firm says.
The company has several safeguards in place to tackle the issue of fake content, which include a Know Your Client process that verifies their identity. The firm is also part of the Coalition for Content Provenance and Authenticity, an organisation which aims to apply watermarking to content.
Launched in 2017, Synthesia originally built dubbing and lip-synching tools and in 2020 switched to launch its first AI avatars for corporate clients. Last year, it reached unicorn status, valuing the company at more than €1 billion.
