# CENOMI × MAPIC 2026: an AI kiosk in French, English, Arabic

Source: https://slash-digital.io/en/case-studies/cenomi-mapic-2026/

CENOMI · MAPIC 2026 An iPad in French, English and Arabic for CENOMI's stand in Cannes. Visitors pick a spot in the mall, Gemini places them in the scene, and they leave with a 2:3 souvenir, printed or on their phone.

Start a project
Watch the video

The Arabic journey, right to left: spot choice, live framing, countdown and generation. Recorded in demo mode, with a drawn mannequin in front of the camera.

In short
For CENOMI's stand at MAPIC 2026 (Cannes, 3–4 November) we built the kiosk edition of Peoplesnap. It is an iPad Pro in portrait that places each visitor in one of three spots at Jeddah Park, a CENOMI shopping mall. Gemini puts them in the scene, the server composes a 2:3 souvenir with the brand band, and the visitor prints it on the stand or takes it home on their phone through a QR code.

Client: CENOMI · MAPIC 2026

Sector: Retail real estate · trade show

Where: Cannes, France

Year: 2026

Deliverables:
- iPad kiosk in FR, EN and Arabic
- Live preview with MediaPipe
- Two-stage Gemini pipeline
- DNP printing via Raspberry Pi
- QR download and operator panel

Services:
- Interactive experiences
- Artificial intelligence
- Computer vision

3, kiosk languages: French, English and Arabic

3, Jeddah Park spots to choose from

15.3 s, median generation time in pre-event testing

01 The context

## A Jeddah Park postcard, in Cannes

MAPIC is a retail real-estate trade show held at the Palais des Festivals in Cannes; for 2026 the organiser expects about 4,000 participants from 75 countries. CENOMI wanted every visitor to its stand to leave with a postcard of one of its malls with themselves in it, in under 60 seconds and with nobody needed to explain the process. The brief set six linear screens, exact French copy and a 4×6 in output, on paper or on the phone.

02 Design and build

## From the chosen spot to the print

The visitor taps COMMENCER, picks one of the three spots and sees themselves inside the mall before the photo is taken. That cut-out runs on the iPad with MediaPipe, and nothing is uploaded. A framing assistant turns the frame green, and a «Regardez ici» cue draws the eye to the camera, which sits on the side bezel of an iPad Pro in portrait. After the 3-2-1 countdown, the photo goes to the server.

Generation has two stages. Gemini (gemini-3.1-flash-image) edits the photo onto a green background. It keeps the participant's face and build, and dresses them in the outfit and CENOMI bags from the client's reference montage. A deterministic step then keys out the green and places the person at the montage model's exact position, over the real photo of the spot. sharp adds the #21113F band and the logo; the AI never draws the logo. In pre-event testing, with 8 synthetic subjects across the 3 spots, 24 of 24 images came out, with a median of 15.3 s.

For printing, a Python agent on a Raspberry Pi pulls jobs from the server and sends them through CUPS to a DNP QW-410, which prints a 4×6 in about 19 s. For the phone, the kiosk shows a QR code and confirms «Photo récupérée» once the visitor opens the link. It then counts down 5 seconds and returns to the start for the next person.

-
01

### Three languages, one of them right to left

The French is the brief's copy, word for word. English and Arabic are picked on the home screen, and in Arabic the whole interface flips. The phone page follows the browser's language. The Arabic copy is still awaiting CENOMI's sign-off.

-
02

### Visitors see themselves in the mall before the photo

MediaPipe separates the person from the background on the iPad itself and shows them inside the chosen spot in real time. The assistant asks them to step into the frame or hold still, and the image stays on the device until capture.

-
03

### Printing with real status

The Raspberry Pi polls the server, prints, and reports whether a job is queued, printing or printed. The panel shows the prints left in each 150-print kit, and if the printer fails, the kiosk switches straight to the QR code.

-
04

### A hidden panel for the stand team

Behind a PIN, the team sees the status of the AI, storage and printer, today's counters, and the controls to pause or reprint. The counters leave test sessions out, so the event figures stay clean.

03 Gallery

## The project in pictures

7 images · Open any image to see it full size.

The three Jeddah Park spots, with the French copy exactly as the brief wrote it.

Live preview: the visitor already stands inside the mall, and the frame turns green once the framing is right. A drawn mannequin stands in for the person here.

The same screen in Arabic, with the whole interface mirrored right to left.

The result screen in Arabic: print, get it on your phone, share or start over. The picture is the demo-mode composite, not the Gemini output.

The phone download page in English: download, share, the AI-generated label and the deletion date, 30 days out.

The same page in Arabic: the language comes from the phone's browser.

The operator panel, captured in demo mode: status of the experience, the AI, storage and the printer.

04 Outcomes

## Ready for 3 November

MAPIC has not opened yet, so there are no usage figures. The kiosk counts sessions, prints and QR scans, and we will build CENOMI's report from them; the numbers and stand photos will be published here after 4 November. The code was written between 20 and 22 October 2026 and reuses about 60% of the Nigloland 2026 backend.

- iPad kiosk in French, English and Arabic (RTL)
- 2:3 souvenir printed on the stand or downloaded by QR
- Session, print and QR counters for CENOMI's report
- Still to come: real figures and stand photos after the event
