One-Way and Two-Way Avatars: Why Video Avatars Are Not Roleplay

After my post on the three levels of AI roleplay, most people got the message: conversational avatars are nothing like Copilot, and they're not a chatbot with a static face on it either.

The follow-up question is often: “So it's similar to video avatar tools?” A fair guess, but no. These are two different technologies, built for two completely different purposes.

One-way avatars deliver content

With video avatar tools, you write a script, it's turned into speech, lip-synced onto a face and exported as a video. Once it's rendered, it's done: the same video, every time, for every viewer. The purpose is to deliver content, such as onboarding videos, announcements and product explainers. You watch it, and that's the whole interaction.

Two-way avatars let you practise

The other kind is what I call a two-way avatar, because the talking goes both ways. That's what we build. It's a real-time, unscripted conversation: you talk, it listens and responds live, based on what you actually said. It remembers the conversation, stays in character (a tough customer, a nervous first-time manager, a candidate in an interview) and reacts to you in real time.

The purpose here is practice: rehearsing a real conversation before it happens and building soft skills.

Why the difference matters

One delivers information. The other lets you practise in a safe space, get it wrong, adjust and try again, which is what builds behavioural change. A video will never do that, no matter how many times you watch it.