Generative AI has captivated the public by creating intricate, seemingly authentic text and images from verbal cues. However, closer inspection often reveals imperfections in the results.
Observers have noticed oddities like strange fingers, disappearing floor tiles, and mathematical errors.
Now, Synthesia, an ambitious AI startup specializing in video creation, especially customized avatars for businesses, is introducing an update to tackle challenges in its field. The latest version boasts avatars based on real humans for enhanced emotion, improved lip-syncing, and more authentic movements when generating videos from text inputs.
Unlike other generative AI companies like OpenAI, which target both consumer and enterprise markets, Synthesia is solely focused on developing human-like generative video avatars for business applications such as training and marketing.
This focused approach has helped Synthesia carve a niche in a competitive AI market that risks becoming commoditized once initial hype subsides.
Synthesia’s latest release features Expressive Avatars, described as “the world’s first avatars fully generated with AI.” These avatars are based on large, pre-trained models to emulate human speech patterns more realistically.
CEO Victor Riparbelli acknowledges that there is still room for improvement in the technology, particularly in capturing intricate human details like facial expressions and hand movements, which are notoriously challenging.
Despite the challenges, Synthesia continues to grow, with a valuation of $1 billion and ongoing efforts to enhance AI safety and prevent misuse of its technology.
The company’s latest version aims to attract more users and expand its reach in creating AI-generated video content across multiple languages for a variety of high-profile clients.
