──── BLOG |

Digital human and avatar technology: how one is actually built

Behind a convincing digital human there is no single technology. There is a chain: a likeness, a rig, a face that can speak, a brain of some kind, and an engine that runs it wherever it has to run.

This guide walks that chain in plain language, so you can tell which part a quote is actually covering, and which part is where projects overrun.

2. The rig: what makes it move at all

The rig is the skeleton and the controls underneath the model. It is invisible, unglamorous and the single most common place a digital human project goes wrong, because a bad rig cannot be fixed by animating harder.

For the face it is a separate problem: expressions are built as blend shapes — a library of poses the face can move between. How many you build sets the ceiling on what the character can express, and that decision is made at the start, not later.

4. The behaviour: from a script to a conversation

At the simplest level the figure plays a sequence. One step up, it follows a decision tree: the user chooses, the character responds, and every branch was written and approved beforehand.

At the top, it answers freely. That is where our conversational assistants sit — answering in real time, collecting data from each interaction and guiding the user through scientific or commercial information. In health that freedom is the expensive part, because every possible answer is a claim somebody has to stand behind.

Where the money actually goes

Rarely on the model. It goes on the face rig, on validation of what the character is allowed to say, on lip sync per language, and on the number of channels. A project that decides those four things in the brief costs a fraction of one that discovers them halfway through.

PROJECTS

• Corsa
• Bionet

RELEVANT SOLUTIONS