We are onboarding the first creators one by one, so every twin sounds right before anyone talks to it. Leave your email. We open access in small batches.
No spam, no charge. We write to you when it's your turn. How we handle your data.
Already sure? Book a call:
Your channel, your podcast episodes, your interviews. What you have already published is enough: nothing to record, no questionnaire, no studio session. That is the only moment we need you, and it takes a few minutes.
We extract the way you speak, your topics, your strong opinions, down to your recurring turns of phrase. Then we test it: does it sound like you, does it dare to disagree, does it hold under pressure. Until it passes, it does not ship.
You talk to it, you tell us what rings false, we fix it. Nothing goes live until you have approved it, and you can have it taken down whenever you want, without explaining why.
Your fans click, subscribe, and message the twin on WhatsApp. There is nothing left for you to do, and you get paid on every conversation.
Nothing. Building the twin is free, so is putting it live, and there is no subscription. We only earn on what your twin earns: fans pay a monthly subscription, and you keep most of it.
If nobody subscribes, it cost you nothing. That is deliberate: we cannot sell you a tool and leave you alone with it, we have to have a stake in your twin being good.
On every screen, and it cannot deny it even if a fan pushes. This is not politeness: the European AI Act requires it, and it is what separates a consented twin from a deepfake. Our guide on the legal framework for AI twins covers it in detail, in French.
Moderation is written in code rather than in its instructions, so it cannot be talked around. A fan in distress gets a real helpline before any other answer, subscriber or not.
And that is a choice, not a technical limit. A confidence is only possible if nobody reads it back afterwards. You will know how many fans talk to you and about what, never who said what.
It is the only question that matters, which is why we open in small batches rather than self-service. What decides is the format of your content more than its volume: someone who talks to camera gives a better twin than an interviewer, whose public speech is mostly questions. We tell you before we start if your material is a problem. Details in our guide on how to build an AI twin, in French.
Less than people expect. A few tens of thousands of words are enough for a first version, so a handful of long episodes or interviews.
It will, like any language model. That is why it never gives medical, legal or financial advice, why moderation sits in code, and why you can have it pulled with one message. We would rather tell you now than find out together.
No, never, neither yours nor your fans'. The twin is built from your public content, and nothing else. It is also a contractual obligation from WhatsApp, explained in our guide on what Meta actually allows, in French.
Sounds right? Leave us your email, or book a call. We answer everyone.