How it works The analysis Simulations Dashboard Lead generation via Google Ads Gallery See it for my clinic
AI app for med spas

Analyse their skin. Suggest the treatment. Show them the result.

Lexivo gives your clinic a branded AI skin analysis and treatment simulator. A prospective client uploads one photograph, gets a detailed read of their own skin, sees the treatments that address what it found — and then sees themselves after a course of the leading one. That is when they ask for a consultation.

Up to eleven indicators, marked on their own face
Treatments drawn only from the menu you enable
A simulation generated from their photograph
A skin analysis report: an overall score of 77, a rail of scored indicators, the client's photograph with a hydration overlay, and the reading for hydration
The analysis
A client's own photograph before an IPL photofacial simulation
The same photograph simulated after an IPL photofacial
Their photoTrueRender™
The simulation
Fully branded as your clinic
How it works

One photograph,
and a client who wants to come in.

Five screens, in the order a prospective client meets them. Everything below is the running product, carrying a demo clinic's branding rather than ours.

  1. A clinic-branded home screen carrying the clinic's own name and palette
    Step oneYour app, your name Your clinic's name, palette and treatment menu. Nothing on screen says Lexivo.
  2. The photo step, showing the consent line a client agrees to before analysis
    Step twoOne photograph Taken on a phone, wherever they are. Consent is asked for before anything is analysed.
  3. The skin analysis: an overall score with named indicators and markers on the client's face
    Step threeA deep skin analysis Up to eleven named indicators, scored and marked on their own face — not a number and a mood.
  4. Suggested treatments, each naming the indicators it addresses
    Step fourThe treatments that fit Each one names the indicators it addresses, and every one of them comes from your menu.
  5. A chemical-peel simulation as a drag-to-compare wipe across the client's own photograph, above a button to request a consultation
    Step fiveTheir new look A simulation of what a course of that treatment could look like — on them, and then a consultation request.
The analysis

Eleven indicators — and the machinery to withhold every one of them.

The usual claim in this category is the size of a training set. That number describes a dataset, not the photograph in front of you. Everything below is a property of the pipeline that runs on this face: an ensemble, a quorum, a contrast threshold — each one versioned, so two analyses are only ever compared when they were produced under the same rules.

Each indicator is scored 0–100 and banded, weighted into one composite, and attributed to the region of the face it was read from.

HydrationOilinessBlemishesDark spots RednessTexturePoresRadiance FirmnessFine linesDark circles
The skin analysis screen: a weighted composite score, the eleven-indicator rail, and markers drawn on the client's own photograph
Their own photograph, read
5 → 1 Ensemble per indicator Five independent passes, reduced to their median. Below three usable passes there is no analysis at all — rather than a complete-looking one built from a single sample.
±5 pts Stability window A pass agrees if it lands within five points of the median, and 80% agreement marks an indicator stable. A second band check catches a split ensemble that a low deviation would hide.
9 of 11 Composite quorum The overall score is withheld unless nine indicators were assessable, and weights are renormalised over what was available. Below four, the report itself is withheld.
4 / 3 / 3 Evidence thresholds Minimum luminance, red-excess and crease contrast against local skin before a marker is drawn. A candidate that fails is dropped, never nudged somewhere plausible.
0.4% Probe radius The sampling disc as a fraction of image width — smaller than a freckle on purpose, because a wider disc averages a spot together with the clear skin around it.
1536 px Scoring crop The face is cropped and resampled to this long edge in normalised sRGB, so it is read at full patch density instead of at the fraction of the frame it happened to occupy.
  • An unassessable indicator returns nothingNot a neutral default. The pipeline this replaced returned 72 for anything the model could not see, and that number was then banded, weighted into the composite and drawn on the face with a confident ring.

  • Position is chosen spatially, not by scoreEach indicator's marker is the box with the highest median overlap against the other passes — an existing box, never an average of several disagreeing ones, which would drift toward the middle of the face and match nothing.

  • Image quality is judged separatelyA single pass cannot be generous about a face and strict about a photograph at once. Quality is its own question with its own severity ladder, and a refusal names what to change.

  • Every analysis is stamped with the rules that produced itPrompt, calibration, aggregation, geometry and image handling each carry a version. Two analyses are compared only when those versions are comparable — otherwise a changed threshold would make incomparable reports look like progress.

The full method, in detail

The simulation

Then they see themselves after it.

A treatment menu asks a person to imagine. A simulation of their own face does not. Each pair below is one treatment, generated by TrueRender™ from the photograph on the left. Drag to compare.

A client's own photograph before an IPL photofacial simulation
The same photograph simulated after an IPL photofacial
Their photoTrueRender™
IPL photofacial · uneven tone and visible pigment
A client's own photograph before a chemical peel simulation
The same photograph simulated after a chemical peel
Their photoTrueRender™
Chemical peel · dullness and surface texture
A client's own photograph before a radiofrequency tightening simulation
The same photograph simulated after radiofrequency tightening
Their photoTrueRender™
Radiofrequency tightening · definition along the jaw
A client's own photograph before a microneedling simulation
The same photograph simulated after microneedling
Their photoTrueRender™
Microneedling · surface irregularity and scarring

Simulations are illustrative visualisations, not predictions or promises of treatment results. Individual outcomes vary, and only a qualified practitioner can determine what is appropriate. Where the engine is uncertain it is instructed to under-improve rather than flatter.

Cosmetic only

The analysis describes visible appearance in a photograph. It does not name a skin condition and it does not attribute a cause.

Your clinician decides

It does not judge whether somebody is a candidate, identify what might rule a treatment out, or set device parameters.

Consent and deletion

Consent is asked before anything is analysed, and photographs are removed automatically once the retention window passes.

Lead generation via Google Ads

Every clinic bids on the same few words.
Almost nobody bids on the question underneath.

A treatment page can only be found by somebody who already knows the treatment's name — so that is where the whole market competes at once, over a short list of procedure words. What people actually type is a symptom: dull skin, dark patches, the lines that arrived this year. A menu of treatments is not an answer to a symptom. An analysis is, and it names eleven of them.

  • Your app is the landing pageThe click lands on a read of their own skin, not on a contact form asking somebody who is still deciding to pick up the phone.

  • Eleven indicators, eleven kinds of questionEvery symptom the analysis can name is a search a treatment page has no answer for.

  • Filtered to the treatments you offerNothing is suggested that you do not sell, and the shortlist is decided by the scores alone — not by a second guess from a model.

  • The click travels with the enquiryThe identifier Google Ads puts on a click, and the campaign behind it, arrive attached to the enquiry rather than stopping at the landing page.

What they typed, and where it lands

Six of eleven
What they typedIndicator
why does my skin look tiredRadiance
what are these dark patchesDark spots
how do I get rid of acne scarsTexture
my skin is dry whatever I useHydration
the lines that arrived this yearFine lines
my pores look bigger than they werePores

Six of the eleven. This is what the analysis scores, not a claim about how often anybody searches for it.

What arrives with the enquiry

On
gclidutm_campaignutm_sourceutm_term

Captured the moment somebody lands, held for the visit, and attached to the enquiry when they ask to be contacted — so the campaign that paid for the click is still on the record at the other end.

Keyword lists and campaign work are the separate Lexivo marketing product that ships alongside the app on one login — you run it yourself, or we run it with you. The app's part is what happens after the click.

App marketing dashboard

The app is the offer.
The dashboard is the machine behind it

An app nobody opens is a brochure. So it ships with a dashboard that fills it: send the analysis to every enquiry you already have, point ads at a free skin analysis instead of at a treatment page, let the follow-up run while you are with a client, and keep an honest record of which enquiry turned into a consultation. The three panels below are screens of that dashboard.

Send the analysis to every enquiry you already have

Send now
318 selectedEmail · Text
  • Amara K.Uneven tone · enquired in March
  • Ines M.Fine lines · consultation, never treated
  • Rosalie D.Acne scarring · treated over a year ago
+ 315 more

Import a spreadsheet, or use the enquiries already in there. Every past enquiry that went quiet gets a link to analyse their own skin — with no advertising spend attached to it.

What they searched before they found you

Last 30 days
SearchAnalysesEnquiries
skin analysis near me38647
pigmentation treatment27139
acne scar treatment18826
skin tightening15518

Keyword lists and budgets are built around the treatments you actually offer, then handed to you or your agency as one file.

Follow-up, running on its own

On
  • Messages sent
    212
  • Opened
    148
  • Ran an analysis
    61
  • Consultation booked
    39

People who opened get a soft nudge. People who ran an analysis get a different one. Both stop the moment somebody books.

Ads that point at the analysis

Not at a treatment page. Somebody months away from booking anything will not fill in a consultation form, but they will analyse their own skin.

Email and text sequences

Sent in business hours in their own timezone, from your authenticated address, until they book or ask you to stop.

It watches what they did

Opened but did not click, analysed but did not book, simulated a treatment twice and vanished — each gets a different next message.

Every enquiry keeps its context

The analysis they ran, the treatments they opened and the simulation they generated, attached to the enquiry so nobody starts the call cold.

Measured on booked consultations

Not clicks and not form fills. Every enquiry keeps the ad, post or search that produced it, through to the consultation.

Opt-outs handled properly

Honoured everywhere, straight away, on every channel, and logged in case anybody ever asks you to prove it.

Figures shown are illustrative — they show the shape of the report, not a rate anyone should expect. The dashboard is a separate Lexivo product that ships alongside the app on one login: you run it yourself, or we run it with you.

Questions

The ones clinics ask first.

No. It describes how skin appears in a photograph and routes to treatments your clinic has enabled. It does not name a condition, and the consent line a client agrees to before analysis says exactly that, in plain words.

Yes. The treatment menu is a setting in your admin area, and the client-facing suggestions are filtered to it. A treatment you switch off never appears. The one thing that can be suggested without being something you sell is general skincare advice — a routine, or daily sun protection — which is deliberate: an app that only ever sells is one clients stop trusting.

The client is asked for another photograph and told what to change — the specific reason is carried through to the message they see, rather than one generic failure. Where a single indicator is obscured, the rest of the face is still analysed and that indicator is reported as unavailable.

No. It runs in a browser on your own address, so it can be linked from a page, an advertisement or a message. There is nothing for a client to install.

Consent is asked before anything is analysed. Uploads are deleted automatically once the retention window passes — ninety days on the skin configuration — by a job that runs on a schedule, and the deploy checks that the job survived the release rather than assuming it. Full detail is on the privacy page.

The consultation button points at your own booking page — a scheduling link from the usual providers, or any address you set. Enquiries also land in the Lexivo enquiry inbox that ships alongside the app, carrying what the client explored and asked. There is no deeper practice-management integration today; if that matters to you, say so on the call.

Request a demo

Run it on a photograph you bring.

Fifteen minutes. We analyse a real photograph while you watch, simulate a treatment from your own menu, and show you the enquiry it produces. No slides.

  • Watch an analysis run liveIncluding the parts where it declines to answer, which is the part worth seeing.

  • See a simulation of a treatment you offerYour menu, on a real face, at the quality a client would get.

  • See your clinic's branding on all of itYour name, your palette and your next step, as a client would meet them.

We reply within one business day.

Thank you

We'll be in touch within one business day.

See Live App