Chen-Ping Yu

After Google acquired my AR startup, I started thinking about the next interface, this time for AI. Text and voice work fine, just like DOS did before the Mac. But nothing’s more intuitive than face to face.

After two years of research, today we’re launching Atmee v2: a real-time avatar model that makes live video agents feel truly life-like.

• Any character from 1 image + 10s of audio
• Contextual expressions, gestures, even interacting with the environment
• Live video calls, voice only, or video generation

At its core, Atmee v2 is a video foundation model, it can render anything, but it’s built to excel at characters, which is what video models like Seedance 2.5 and Wan 3.0 are mostly used for. In comparison:
Atmee v2

• Up to 50x faster
• 10-100x cheaper

Build tutors, sales coaches, support agents, companions, game characters, even whole living worlds. The interface after text and voice starts today.

The Jev & DeepSeek moment for video generation, with Atmee v2 real-time avatar model.

119x cheaper and 51x faster than Seedance 2.5, one of the best video models out there today. And the quality? See for yourself.

Why this comparison? Because Atmee v2 is fundamentally a video foundation model. iIt can generate video just like Wan, Seedance, and MiniMax. So we ran a side-by-side on an 8-second clip, and the result was just... wow.

API launching soon, sign up here for early access: t.co/WptcekAtKT

Forcing Atmee v2 (our live avatar model) into generating a general scene: a poor man’s world model.

Watch the guy take a stroll, go order a bowl of noodles, and enjoy it, and interacting with a world that’s being generated live.

Sure, the quality is still lacking, but this is an unintended use of an avatar-centric model, running on a single RTX-grade GPU, streaming and generating in real time. So fun.

Want to build your living world & live avatars? Sign up here for API early access: atmee.ai/v2signup

Have you ever wanted to video call with your favorite movie character?

I was watching the movie Obsession, and really loved the story and thought how Nikki was portrayed so well with the different emotions and micro expressions, and the erratic/psychotic personality.

Built a Nikki with our Atmee v2 model, I thought the result blew me away.. I felt I was really talking to her, and I was kind of scared by her behavior.

Just a quick test but wow the possibilities are endless on the interactive entertainment side.

API is launching soon, sign up for early-access here: lnkd.in/dtnBXFDt

Streaming a world model that generates chunk by chunk on a single GPU. Fidelity can be better as well as scene consistency, but it gets a big thumb up for efficiency.