NVIDIA-Backed Startup Showcases Super Realistic AI Avatars With Human Emotions
Synthesia, a startup that has the backing of NVIDIA, is launching its fourth generation of AI avatars it’s calling Expressive Avatars that are powered by the EXPRESS-1 model. These avatars are designed to be used in video production and deliver what the company says are “dynamic and lifelike digital personas that blend the best of human and artificial intelligence into one seamless experience.”
The model these new AI avatars use is built to be able to translate the words in a script into a result that matches. It’s an improvement on prior models where an avatar might deliver a dead pan result when the words coming from its mouth indicated any range of emotions. According to the company, avatars generated with this new model use “correct tone of voice, body language and lip sync, like a real actor would.”
The biggest leap in this new version of AI avatars is that it no longer relies on pre-recorded or ore-defined dynamics. Instead, the new model allows for the system to generate results on the fly. Synthesia says that its “capable of generating entirely new and unique performances, greatly expanding the range of expression and interaction.” Users can also easily regenerate a take if a particular result isn’t working out.
Synthesia is also tackling the safety issues that might arise with the release of this new version of AI avatars. The company acknowledges that it software might be misused by bad actors to inflict harm. To help combat this Synthesia says it’s “investing in the early detection of bad faith actors, increasing the teams that work on AI safety, and experimenting with content credentials technologies such as C2PA.”
What do you think of these new AI avatars? Are they convincing enough to believe that they’re actually humans? Let us know in the comments below!
