Service 03 · Real-Time AI Avatars

Real-time AI avatars — a face for the machine.

Chat is fine for most things. But some experiences need a face — a guide, a greeter, a coach that looks back at you. Lifelike heads, live voice and a personality, streamed in real time and answering in the moment.

Where this one actually stands

This is frontier work, and we say so. Avatar builds here are proven concept work with client projects underway — not a shelf product with a decade of deployments behind it. If you want a supplier who will tell you it is all solved, we are the wrong lab. If you want one who has already built it once and knows exactly which parts bite, keep reading.

Why A Face

Some jobs a text box simply cannot do.

Text is efficient, and for most business problems it is the right answer. But there are moments where the interface being a person matters — where someone needs to be walked through something, welcomed, taught, calmed, or held long enough to finish.

A face changes the register of the interaction. People explain themselves more fully to something that appears to be listening. They stay longer. They ask the follow-up question they would not have typed. That is not a gimmick; it is the same reason a shop has staff rather than a printed sign.

The engineering question is whether the illusion holds. An avatar that pauses a beat too long stops being a person and becomes an uncomfortable video call with a machine — and once it breaks, it does not come back within the session.

What Goes Into One

Six things that all have to be right at once.

Live presence

Streamed video of a face that responds now — not a rendered clip fetched after the fact.

Voice

Speech that carries the right personality, pace and accent, and does not sound like a station announcement.

Lip-sync

Mouth and words agreeing frame by frame. This is the detail people cannot articulate but instantly notice.

Latency budget

Everything above, inside the window where a pause still reads as thinking rather than as broken.

A brain worth talking to

Underneath the face, an agent that actually knows your business — otherwise it is a beautiful head saying nothing.

Disclosure

People are told they are speaking with an AI. Non-negotiable, and increasingly the law as well as the decent thing.

The Strange Part

Latency is the uncanny valley’s bouncer.

Every stage costs milliseconds: hearing the speech, understanding it, deciding the reply, generating the voice, driving the face, pushing the video down the wire. Individually all of them are fine. Added together, they are the entire product.

So the build is a budget, not a feature list. Every millisecond is a design decision, and the honest trade-offs — a slightly simpler face for a faster reply, a shorter answer to start speaking sooner — get made deliberately and explained, rather than discovered by a client in a demo.

The avatar build is written up in full on the case studies page, including where it currently stands and what remains hard.

Questions

The ones we actually get asked.

Can I have a real-time AI avatar on my website today?
A conversation about one, yes. This is badged as frontier work: the concept builds are proven and client projects are underway, so a scoped project starts from something already built once rather than from a blank page. If your timeline needs a fully mature, off-the-shelf product, we will say so at the scan rather than after a purchase order.
What can a real-time avatar actually be used for?
The useful cases are the ones where a face earns its place: a guide that walks someone through a complicated process, a greeter that qualifies and routes, a coach or trainer that holds attention, a spokesperson for a product that needs explaining. If a text box would serve just as well, we will tell you to use the text box.
Why does latency matter so much?
Because the illusion is time-based. A reply that arrives a beat late stops reading as a person thinking and starts reading as a machine struggling, and once a user notices, the effect does not recover within that session. Every stage of the pipeline is therefore budgeted in milliseconds from the start.
Will people be told they are talking to an AI?
Yes, always. Disclosure is built in, not bolted on. It is the right thing to do and it is increasingly a legal requirement — our position on AI transparency is published on this site.
Can the avatar be connected to our own systems?
That is the point of doing it here rather than buying a talking head. The face is the interface; underneath it sits a custom agent connected to your data and systems, which is the same work described on the AI agents page.
The Rest of the Lab

These combine.

Most builds use two or three of them together. See all services.

Step One

Start with one honest sentence.

“Here’s what’s quietly eating my week.” Tell Ray in your own words. A clear, priced build path comes back, and a human signs off everything that leaves the lab.

Start the scan — free
Free scan · Fixed prices · You own everything