A Work In Progress:

We’re carefully considering the best way to bring this training material to life in a repeatable way, that can be accessed remotely on a virtual headset.

Key considerations and challenges:

Avatars and Realism

It felt natural to use realistic avatars to try and make more empathetic human characters. But, this raises the bar for realism all over – users expect the faces to animate in a realistic manner too – something we didn’t want to invest in at the prototype stage.
We thought we could make do with off-the-shelf animations for early iterations and later move to motion capture and advanced animation to bring the experience to life. But, when we tested our first prototype with users, the avatars and their animation was all people could talk about.

We’ve changed to use Ready Player Me avatars, whose stylised, cartoonish looks fit better with simple animations.

Natural animation

Something not on display in this version unfortunately is a work in progress project we’re developing for procedural natural facial animation. In it’s current version we left it out and you’ll see very little facial animation – just the standard lip sync that the avatars come with. We’d love to collaborate on tools in this area, so please reach out if you’d like to talk machine learning drive animation.

To AI or not to AI?

It is tempting to use the natural language capabilities of AI and include avatars that don’t just speak their lines, but allow interaction and natural responses. However, when we consider what we are trying to create, which is a scripted, and tightly controlled experience, it just doesn’t fit.

An AI avatar might impress users and feel nicer to interact with naturally as opposed to choosing lines – but also enables off-script interaction which can dilute or even garble the messaging.