Sean Parker’s social app has traded the empty-room problem for a cast of AI characters. Now Cantina Labs wants those characters to talk, perform, and follow their creators onto the rest of the internet.
LemonSlice is a San Francisco AI research and product lab that turns a single photo into a real-time, talking video avatar - the visual face for voice agents and chatbots. Its Lemon Slice-2 model is a purpose-built, large-scale video diffusion transformer that generates every pixel from scratch to produce expressive, full-emotion talking characters (human or cartoon) that livestream at conversational speed. Founded in 2024 by three PhD-creators (formerly Infinity AI), the company raised a $10.5M seed led by Matrix Partners and Y Combinator to make all video interactive.