Meet Essence 2 and Expression 2
Our second-generation models are here. Essence 2 turns one portrait into a photoreal AI character; Expression 2 brings any character to life from one image.
Today we are introducing two new models for real-time AI characters: Essence 2 and Expression 2. Both turn speech into a lip-synced, expressive face as it is spoken, so your app, agent or website can talk to people face to face.
Essence 2: a photoreal person from one portrait
Give Essence 2 one portrait and it makes a lifelike AI character that speaks, listens and reacts in real time. It is made for people: guides, presenters, tutors and support staff your users can talk to.
Expression 2: any character, animated live
Expression 2 brings anything with a face to life from a single image: a person, an animal, a mascot or an object. It is built for brand characters and storytelling, where personality matters as much as realism.
Where they run
Both models run where your users are:
- On the device: iPhone, Android and Mac, and even a Linux PC with no GPU.
- On your servers, for products that keep audio and video on their own hardware.
- In our cloud, the quickest way to start.
When the character renders on your hardware, its audio and video stay there.
Try them
Talk to characters made with them on Explore. Make your own from a photo in the Console, or start building with the developer guides.
Our first-generation models, Essence 1 and Expression 1, stay supported.