AI digital clone: how they work and what they cost

An AI digital clone answers as one named person, from that person's own material, in text, in their voice, and as a live talking avatar. On AIMayDo the build is free, there is no subscription, and you set the per-minute price yourself. What decides whether it is any good is the material you give it, and we publish a measured score for how closely it tracks you.

The offer

A verified clone of you, which we call a twin, live on the internet, answering in text chat, in your own voice, and on a live talking avatar. One build gives you all three. It runs on your material, in your manner, at the hours you are not available, and it carries your name because your identity is checked before it goes live.

Building it is free. There is no subscription, so a quiet month costs you nothing. You set your own per-minute price above a platform base and keep 80% of your markup, with commission on the markup only. You choose what a first-time visitor gets free. You approve what it may say before it says anything, and you can switch it off at any moment.

Going live needs no plan at all. Consent and a checked identity are the only gate. Paid seats add the things you want once it is working: embedding your twin on your own site, keeping it out of the public directory, usage analytics and response controls, and an allowance of sponsored minutes to give your visitors.

What a digital clone actually is

The phrase covers three separate layers, and they are not equally load bearing.

The knowledge layer is a language model working from your writing, your recorded answers and your published positions, together with a description of how you argue. It is what lets the clone handle a question it has never seen in something close to your reasoning. Everything above it is presentation.

The voice layer is a model conditioned on recordings of you, able to speak any text in your voice. On its own it knows nothing. It repeats whatever is fed to it, so its quality is borrowed from the layer underneath.

The face layer is your likeness animated in time with that voice, so the thing looks back at a visitor while it talks. A face raises an expectation that the answer then has to meet, which is why we treat the material as the part that decides whether a twin is any good.

On AIMayDo the three come from one build. You supply your material once, and the same twin answers in text chat, speaks in your voice, and appears as a live talking avatar. There is no second setup for the second modality and no third for the third.

How the money works, in five lines

Building a twin is free.

There is no subscription, and nothing is billed in a quiet month.

You set your own per-minute price above a platform base.

You keep 80% of your markup, and commission applies to the markup only, never to the base.

Visitors are billed only while they are actually talking, and you can revoke the twin whenever you want.

The longer version, including what a running meter does to a conversation and how to think about setting your number, is on charging people to talk to your AI clone.

How one gets built

Strip out the marketing and the pipeline is short.

You collect source material. Long-form writing, talk transcripts, question-and-answer sessions, case discussions, anything where you took a position and defended it. The most useful thing you can hand over is a large set of real questions with your real answers, including the ones where the answer is "it depends" and you spell out what it depends on. A bio and a handful of posts give the model very little to work from.

You record consent, in your own voice, agreeing to be represented this way.

You verify who you are. Stripe checks your identity before the twin goes live, so the name on a twin is a name somebody stood behind.

You approve the decision rules. We extract the rules your material implies, the way you weigh one consideration against another, and you read them and approve them before anything is published. Nothing goes out under your name that you have not signed off first.

After that the material is prepared so the model can retrieve from it, the voice is fitted, and the face is animated over the voice. The mechanical steps are quick. The material is the slow part, because it is the part only you can supply.

What it really costs

"What does an AI clone cost" has a misleading answer, because the fees are the smallest item on the list.

The first cost is material. A clone compresses what you have already said and written. Years of published answers and recorded argument give it something to compress, and thin material gives it very little. No amount of processing invents the difference.

The second cost is your own review time. Someone has to read what the thing says and judge whether it is right, and that someone is you, because nobody else can tell. Your positions move over time and the material you uploaded does not move with them, so treat it as ongoing work.

The third cost is exposure. Something carrying your name and face will produce answers you never personally wrote, and some of them will be wrong. In fields where being wrong has consequences, medicine and law and finance among them, that exposure is real, and a disclaimer at the foot of a page does not remove it. The things that help are the dull, checkable ones: identity verified before launch, decision rules you approved, a published measure of how close the twin is, and a switch that takes it down straight away.

What we publish on every twin

Fidelity gets measured by our own engine, and the measurement is published. That engine builds the twin's profile across five dimensions, voice and tone, reasoning style, lexicon, positions and guardrails, each carrying its own confidence, and it drafts the decision rules the expert approves before anything goes live. To measure it we hold back a set of questions the expert answered themselves, put those questions to the finished twin cold, and score the answers against a rubric we wrote. The score sits on the twin's page, where a visitor sees it before paying for anything.

We do not describe a twin as indistinguishable from the person, and we would not believe anyone who did. A twin can be confidently wrong at the edges of what you have written down. The score exists so a visitor can see roughly where those edges are instead of guessing. You can read one on Dr Evangelos Katsioulis's twin.

Where AIMayDo is today

AIMayDo is live and running. Dr Evangelos Katsioulis, a psychiatrist and a co-founder, has his twin on the platform now, with its measured fidelity score published on the page. You can talk to it before you decide anything.

We are not going to publish what a twin earns, because we have no evidence that would make such a figure true, and anything we printed would be a guess wearing the clothes of data. Apply the same suspicion to everyone in this category, including us.

Your digital clone on your own site

We are not promising to send you visitors. The part that works today without that is the embed. A twin runs on your own domain through an allowlisted iframe, so it answers people on the page where they already are, who already know your name. Visitors do not subscribe to you. They pay per minute, only while talking, and the meter stops when the conversation does.

You also decide what a stranger gets before any of that begins. The free taster is set per twin by the expert, as two independent caps: an amount of free time and a number of free questions. Either can be set to zero, and whichever runs out first ends the free tier. A twin that gives long consultative answers might be set to a small number of questions with no clock at all. A quick-answer twin might be set the other way round. You set both numbers, and you can change them.

What an AI digital clone cannot do

A twin is not the person. It answers from the material it was given, and it will sometimes be wrong, including confidently wrong at the edges of what you have written down. That is a property of the method, and no version of it ships with a guarantee of correctness.

It is only as good as what you feed it, and feeding it well takes real work.

It does not take you out of the loop. The review work recurs.

And it stays revocable for a reason. If you leave, your face and voice are deleted at the underlying vendors too, not only in our database. That is the control that still works after something has gone wrong in public.

Reasons people build one

Wanting a digital clone because it is a genuinely interesting thing to own is a good enough reason, and we would rather say so than pretend everybody needs a business case first.

Some people want to hear what a version of themselves trained on their own words sounds like. Some want their ideas to answer a stranger at two in the morning, when they are asleep and the question is urgent to the person asking it. Some want to be early in a category that will have a great many people in it later. Some have answered the same question many times over and would like it answered once more without them.

If you want to build one, the build is free and you can take it down whenever you like. Start at build your own verified twin. If you would rather judge the output first, spend a few minutes with the live twin of Dr Evangelos Katsioulis and come back when you have seen what it does.

Questions people ask

What is the difference between an AI digital clone and a general chatbot?
A general chatbot answers as an assistant with no particular identity. A digital clone answers from one named person's own material and in their manner, and on AIMayDo it answers under a verified identity with the decision rules that person approved. It is one specific person's judgement, published so someone else can reach it.
What does it cost to build an AI digital clone?
On AIMayDo the build is free and there is no subscription, so nothing is billed in a quiet month. You set your own per-minute price above a platform base and keep 80% of your markup, with commission on the markup only. The real costs are the material you already have, the time you spend reading what the clone says and correcting it, and the exposure of your name on answers you did not personally write.
Is a talking avatar better than a text clone?
It depends on the knowledge layer underneath, because a face raises an expectation that the answer then has to meet. On AIMayDo you get text, voice and a live avatar from one build, so the question is where to put your effort, and the answer is the material.
How accurate is an AI digital clone?
It is not the person and it will sometimes be wrong, including confidently wrong at the edges of what you have written down. AIMayDo runs its own fidelity engine: it holds back questions the expert answered themselves, puts them to the twin cold, and scores the answers against a rubric we wrote, averaged over several runs. That score is published on the twin so a visitor can see it before paying.
What happens if I want to remove my clone?
You revoke it and it stops answering. Leaving also deletes your face and voice at the underlying vendors, not only in our database. Because the build is free and there is no subscription, nothing carries on costing you after you take it down.

If you want to build one, it is free to start. If you would rather see a live one first, our co-founder Dr. Evangelos Katsioulis, a psychiatrist, has his own twin running.

How it works for expertsTalk to a live twin