Browser-native conversational avatar
Twimi puts on your website someone who listens to that question, answers out loud or in writing and knows your products by heart. Try it right now: it's right there, on the right. It's talking about us.
Ask it a question
Listening
Its gaze follows you
A video avatar has to be built frame by frame, on graphics servers, for the entire length of the conversation. That's what you're billed for, by the minute.
Twimi doesn't send video.
See what we send instead →What it changes for you
They don't come from a commercial choice on our part.
They all follow from the same technical decision, made two years ago.
You pay a monthly subscription and a volume of conversations. A chatty visitor doesn't cost you more than a hurried one, and a campaign that works doesn't blow up the bill.
PourquoiThe others bill by the minute because every minute consumes graphics computing on their side. With us, the image is drawn on the visitor's device.
On 3G in a car park, on a five-year-old mobile, the conversation carries on: the avatar switches to simplified animation, then voice only, then text — without ever losing the thread of what was said.
PourquoiA video that won't load is a conversation that stops. We transmit only voice and parameters: it takes very little to keep going.
Understanding, your documents, the voice and the face arrive together. You have nothing to assemble, nothing to make talk to each other, a single point of contact.
PourquoiDrawing on our experience, we master the design, the integration and the day-to-day operation.
The question everyone asks us first
Putting an artificial intelligence in front of your own customers means taking the public risk that it says something foolish. It's the first objection we hear, and it's a legitimate one. Here's how it's handled — technically, not with a promise.
Why it's possible
A classic video conversation carries dozens of images per second, built elsewhere. Twimi carries the voice and a handful of values that describe what the face is doing — then it's the visitor's browser that draws it, just as it did at the top of this page. Using less energy, Twimi is designed with the environment in mind.
By voice or keyboard. Their question is transcribed.
From your documents, and from those alone. The avatar doesn't invent what it doesn't know.
A catalogue voice of your choice, or a synthesis of the voice of someone in your company, cloned from a one-minute recording.
We send the movement parameters. The face itself is animated by our team.
A direct consequence: low energy use — and no per-minute billing. Across twenty-two major chatbot and conversational-agent platforms we analysed, none offers an animated human avatar today. Twimi offers a written-and-spoken conversational chatbot with a human appearance.
What you get back
The avatar doesn't just answer in your place. It records the questions asked — and it's the first time you see them phrased exactly as they are, without the filter of a form or a search engine.
Three trades, one engine
Twimi isn't a product range. It's a single engine that you configure: your documents, your tone and your goal make the difference between an avatar that qualifies a request and one that advises a purchase.
The avatar welcomes the visitor, understands their real need, presents the right models and passes an already-qualified request to the sales team.
Arno — car retail
The visitor describes what they're after, the avatar suggests the right items in a few exchanges and shortens the path to the choice. A fast, efficient visit.
Pierre — fashion retail
The avatar answers with the brand's expertise, reassures about the product and points to the right contact when needed.
Hélène — cosmetics
Who it's for
Twimi is most useful when the catalogue is large, the information scattered and the answer always the same.
One brand, dozens of locations, local information to keep up to date. One avatar per outlet, central governance, a site-by-site rollout.
“Do you do that in my area too?”
A technical catalogue, hard to browse on your own. The avatar listens to the need and brings the visitor back to the three right references instead of a hundred.
“Which one do you recommend for this use?”
Programmes, funding, prerequisites. The avatar answers on the real content of the courses and qualifies applications before the interview.
“Am I eligible?”
Going live
We don't promise a timeline before seeing your content. That's exactly what the diagnostic is for: assessing your site's AI-readiness, so you leave with a firm scope, timeline and price.
We review your content, your use cases and your site. You receive a costed scope and a date. The amount is deducted from the setup.
Your documents are made AI-ready so they're perfectly understandable by AI agents. The avatar is created with your team, your influencer, your image and voice.
One line of code on your site, or a QR code on your products. Acceptance testing with you, tone tuning, then go-live. Setup and tailored adjustments.
Pricing
You pay for conversations, not compute time. A conversation is only counted if a visitor asks at least one question.
180 €
per month, excl. VAT
390 €
per month, excl. VAT
790 €
per month, excl. VAT
1,490 €
per month, excl. VAT
Prices exclude VAT. Recommended 12-month commitment, 10% discount for annual payment. Displaying the widget is never billed. Data-compatibility work from 690 € to 2,900 €/site depending on the diagnostic.
Twimi Network: one avatar per brand, central governance, an outlet-by-outlet rollout. 1,990 € setup at head office, then 390 €/month, and 39 €/month per location, with volume discounts.
Legal framework
The obligations that fall on you when you put an artificial intelligence in contact with the public are handled from go-live, not afterwards.
Frequently asked questions
On the corpus you entrust to us: product sheets, catalogue, terms, site pages, internal documents you agree to expose. It builds its answers from these documents and nothing else. This scope is decided with you and can be changed at any time.
It says so. Then it offers the next step: rephrase, point to the right page, or pass the question to your team with the conversation's context. An avatar that makes things up costs you more than one that hands over.
Technically, by what travels over the network: they stream continuously built images, we transmit the voice and the movement parameters. Commercially, by what you pay: a subscription rather than talk time. And practically, by what happens when the network is poor: with them the conversation stops, with us it continues another way.
It depends on the state of your content, and that's precisely what the diagnostic measures. At the end of it, you have a firm date. We'd rather give a date we keep than a sales timeline.
The face can be created from a photo or a video. The voice can be chosen from a catalogue or cloned from a few seconds of recording of someone in your company. In both cases, written authorisation from the person concerned is required — we provide the template.
No. One line of code is enough on most sites, whatever their technology. For use beyond the site — on a product, in a shop window, in a catalogue — a QR code or an NFC chip opens the conversation directly in the browser, with no app to install.
Next step
We show you an avatar in action, we answer your questions, and if the topic is ripe we start with a diagnostic of your content — 290 €, deducted from the setup.
twimi.