Google Unveils Live Avatar AI Model for Conversational Avatars

Google has introduced Live Avatar, an AI model that allows brands to create conversational avatars. The feature integrates near real-time live dialogue by pairing the model with video generation and speech. According to Google, the goal is to give companies 'distinct visual identities to fit their brand.'

The Live Avatar feature creates an experience that listens, sees, and speaks with a dynamic visual persona. Google demonstrates realistic and cartoon-like avatars in videos, which act natural and create multimodal conversations by using facial expressions while looking, listening, and speaking in natural language with the user.

The avatars are backed by Gemini's advanced reasoning, allowing Live Avatar to trigger calls for specific tools and fetch data in the background while continuing an active dialogue. Brands can either use a library of pre-made avatars or customize their own through enterprise allowlisting, a cybersecurity practice that compiles an approved list of applications, email addresses, IP addresses, or websites permitted to run or be accessed on its network.

Google has implemented 'strict safeguards' to respect human identities and keep AI-generated content transparent. Each avatar uses Google's SynthID watermarking technology. The company is positioning the tool as something brands can use to sell their products by creating conversational experiences, rather than traditional ad units.