Meta Introduces Codex Avatars, Life-Like Avatars that can be Created Using a Smartphone
Researchers at Meta have introduced its new AI technology with which anyone can instantly create photorealistic avatars with just the help of a smartphone.
Here’s more about it.
Codec Avatars – Meta’s Life-Like Avatars
With the advancements Meta has been making in the area of photorealistic avatars and other AI developments, Meta has jumped years ahead with the Codec-Avatars, enabling anyone to create more life-like photorealistic avatars.
The company first introduced the Codec-Avatars back in 2019, and these learned, photorealistic face models can be used to scan and create three-dimensional images or avatars of a person, that can be later used in VR devices or for such purposes. Meta used MUGSY, a specialized setup that included 171 high-resolution cameras to create the life-like avatars but with the advancements that have been made, the front-facing depth sensor of a smartphone, like an iPhone with FaceID, is all that is needed to create the avatar.
To create the Codex Avatars, the phone has to be panned across the neutral face and then repeat the process with a series of 65 facial expressions. This creates a life-like avatar that can depict more subtleties like raising the eyebrows, squinting, widening the eyes, or scrunching the nose, which will add more life to them.
Researchers have also added that the scanning process takes about three to three and a half minutes and even if it is titled as “instant”, users will have to wait four another six hours for the avatar to be processed. An AI machine with four high-end GPUs is used to create the avatar and depict it in full detail. Even after the product is made available to the public, it will probably not use the GPU on the user’s device but would rather use the cloud GPU for the process.
Meta hasn’t declared a date when it will make it available to the general public, but will soon be made available.
Meta jumps years ahead with lifelike Avatars.
LLMs can be trained to speak like you.
Text-to-voice can sound like you.
The Simulation is coming. pic.twitter.com/dcU3KSMiGj
— AI Breakfast (@AiBreakfast) April 29, 2023
