AI services

AI for brand campaigns and events

Marvy designs and builds AI experiences that put each customer in the picture: photo booths at events, personalized images in campaigns, AI minigames and campaign chatbots. MILO’s AI minigame drew 82,229 users; Manulife’s campaign produced 37,482 personalized images.

Talk to us See the work

Four services, real cases

What we build with AI

Each service below has live work behind it, and each names the case that proves it.

A guest choosing an outfit on SkinCeuticals Skinlab’s AI photo booth at the Skin In Plump event

01 / 04

AI photo booth

Guests pose, the AI restyles them in your brand’s world, and they print the picture or scan a QR code to keep it. A hand gesture takes the shot, so the guest stays in pose.

SkinCeuticals Skinlab: 450+ AI images in 7 days. Invisalign S.E.A. Forum 2026: 4 booths at a forum of 800+.

Manulife’s result screen: two AI-made “green version” images of the same visitor, by a pool and in a dressing room, ready to pick and share

02 / 04

Personalized AI image campaigns

Every customer gets an image made for them: their portrait restyled into your story, their face swapped into your scene, or words that rhyme with their name. It runs on a microsite or in Messenger, and each picture can leave with the campaign’s frame and hashtag, as Manulife’s did.

Manulife, Choose Green for Health: 21,510 users made 37,482 images, with 17,745 Facebook shares. Also Oatside, Chin-su and BVBank.

MILO A2 Cool Milk’s minigame visual with its prizes

03 / 04

AI minigames

Games that put the player in the picture: a child’s face fitted into a themed frame, a child’s drawing turned into a dancing video, spins and prizes run from a CMS that tracks stock in real time.

MILO A2 Cool Milk on Zalo Mini App: 82,229 users, 65,067 plays, 29,200 prizes. PediaSure: 200,000 website visits.

Chin-su chili sauce’s campaign visual: groups of friends in AI-made party scenes

04 / 04

AI chatbots

Campaign chatbots in Messenger: the user sends a photo, the bot checks its quality, swaps their face into AI-made party scenes, and reminds anyone who forgets to send one.

Chin-su chili sauce, with Mono: 20 party themes, 2+ images each, a face swap in about 45 seconds.

From the case pages

Proof, in numbers

Every figure is copied from its case page, and each one links back to it.

  1. 82,229

    users of MILO A2 Cool Milk’s AI minigame on Zalo Mini App

    50,000+ AI frames · 29,200 prizes · 43,000+ public posts · est. 6.5–7.2M organic reach on Facebook

    The MILO A2 Cool Milk case
  2. 37,482

    personalized images in Manulife’s Choose Green for Health

    21,510 users · 17,745 Facebook shares

    The Manulife case
  3. 200,000

    website visits to PediaSure’s drawings brought to life

    50,000 likes · 8,000 comments · 5,000 shares · 60 children awarded a week

    The PediaSure case
  4. 450+

    AI images from SkinCeuticals Skinlab’s booth in 7 days

    500+ attendees took part

    The SkinCeuticals case
  5. 4

    AI photo booths at Invisalign S.E.A. Forum 2026

    3 templates · 800+ professionals at the forum

    The Invisalign case
  6. 20

    party themes in Chin-su’s Messenger chatbot

    2+ images each · a face swap in about 45 seconds

    The Chin-su case

From Manulife and MILO

How an AI campaign is made

The steps Manulife’s and MILO’s campaigns went through, in order.

  1. BriefWhat each customer makes, where it runs, and what the brand needs to see on every output.

  2. Prompt system and test setA prompt system written for the campaign, run against hundreds of test photos in different light and angles until faces stay true (Manulife).

  3. Brand reviewThe brand reviews the AI scenes before launch (Chin-su), and each picture leaves with the brand’s frame and logo, as Manulife’s did.

  4. LaunchOn a microsite, a Zalo Mini App or Messenger. MILO’s was tested on 20+ device types first.

  5. Live monitoringAfter launch, downloads, popular themes and errors are watched and the prompts tuned in real time (Manulife); fixes ship throughout the campaign (MILO).

Brand safety

Keeping AI on brand and safe

What keeps each output on brand, from the first photo to the shared picture.

  • Prompt rules

    Written for each campaign: tone, light, framing and proportions. MILO’s covered the template’s colours, summer brightness and a child’s proportions.

  • Filters before generation

    Photos likely to come out wrong are turned away first. MILO’s engine filtered side profiles, low-quality angles, animal faces, illustrated images and unsafe content.

  • Brand review before launch

    The brand reviews the AI scenes before the public sees them, as Chin-su’s team did, and the outputs leave with the brand’s frame, logo and hashtag.

Questions brands ask

What does Marvy build with AI?

Four things, each with live cases: AI photo booths for events (SkinCeuticals, Invisalign), personalized AI images and words for campaigns (Manulife, Oatside, Chin-su, BVBank), AI minigames (MILO, PediaSure) and campaign chatbots (Chin-su).

Where do these experiences run?

Where the campaign’s audience already is: a microsite (Manulife, Oatside, BVBank), a Zalo Mini App (MILO), Messenger (Chin-su), a web app (PediaSure) or a booth at the event (SkinCeuticals, Invisalign).

How do you keep AI images on brand?

A prompt system written for the campaign and tested on hundreds of sample photos before launch (Manulife); filters that turn away photos likely to come out wrong (MILO); the brand’s review of the AI scenes (Chin-su); and the brand’s frame, logo and hashtag on the outputs.

Do you stay on after launch?

Yes. For Manulife, Marvy tracked downloads, popular themes and output errors after launch and tuned the prompts in real time. For MILO, traffic was monitored live and fixes shipped throughout the campaign.

Start a brief

Talk to us

Tell us what you are planning: the campaign or the event, the dates and the market.

Planning an event? The AI photo booth