In 2017, Synthesia was established to create AI-generated representations of real human faces, such as that of former footballer David Beckham, paired with dubbed voices in various languages. By 2020, the company expanded its offerings, enabling clients to produce professional presentation videos featuring AI avatars of staff or approved actors. However, initial versions of this technology exhibited limitations, including unnatural body movements, inconsistent accents, and mismatched emotional cues.
Recent updates have enhanced the avatars significantly, resulting in more natural movements and voices that better reflect the speaker’s accent, aiming to create a more human-like presentation experience. These improved avatars are expected to serve corporate clients in delivering financial reports, internal communications, and training videos more effectively.
One individual noted that a demonstration of their avatar was both impressive and somewhat unsettling, illustrating the increasing challenge of distinguishing artificial entities from real people. The potential future capability of these avatars to engage in two-way conversations raises questions about their development and the implications of human interactions with AI representations.
When creating an avatar, a prospective user undergoes a calibration process requiring expressive speech and gestures to help construct the AI model. While the process has been streamlined since earlier versions, it still necessitates adherence to specific scripts to generate the desired emotional effect in the avatar’s performance. The resulting videos, featuring the user as a synthetic figure, demonstrate notable advancements in realism compared to previous iterations.
The latest version, known as Express-2, aims to enhance lifelike qualities, with users reporting improvements in aspects like synchronization of movement and facial expressions. By contrast, earlier models faced criticism for issues such as limited emotional range and awkward body movements. Observations indicate a marked improvement in the newer avatars, raising questions about future enhancements and their potential effects on viewer perception.
Source: https://www.technologyreview.com/2025/09/04/1123054/synthesias-ai-clones-are-more-expressive-than-ever-soon-theyll-be-able-to-talk-back/