120k installs were bought: a small Vietnamese team's AI portrait app; the real product is the ad-buying back end

Patrix is an Android app on Google Play where you upload a selfie and generate AI portraits from templates. Two months after launch it hit 125k installs. What drove it wasn't the product, which is nearly identical to dozens of competitors, but the Google Ads buying, accounting, and budget-tuning back end built by its Vietnamese publisher. This piece explains where the volume came from, how the money came back, why you shouldn't copy it by building an AI portrait app, and which part is actually worth learning.

A one-star review with 50 likes

In 2026 Oct 10, a user named David Daniels left a one-star review on Patrix's store page: he uploaded a photo of an older Black woman, chose the 'General' template, and got a young white general. 'Looks a bit like her face, but definitely not the person I uploaded,' and the last line was 'not worth waterfall of commericals', meaning not worth the cascade of ads. By 11-06, this review had 50 'helpful' votes, and it is the most substantive review the app has (Google Play store page, Nigeria storefront, 11-06).

The two highly upvoted reviews beside it tell the story: one person pasted in the Gboard clipboard usage instructions verbatim (7 upvotes), and another started with 'send me bundle please' and trailed off into nonsense (69 upvotes). People who write reviews like this probably didn't search the store for 'AI portrait' and find it themselves. They most likely saw an ad in another app, tapped it, and installed.

What it is

A template-based AI portrait app: pick a theme (football, summer, anime, luxury, art), upload a selfie, and it generates an image of 'you' in that scene. It also offers enhancement for old, blurry, or dark photos, AI object removal, and a works gallery. The store page is labeled 'Contains ads', with no in-app purchases and no subscription. All the money comes from ads.

The developer on Play is listed as Mee Game Inc, registered at an address in a commune in Ho Chi Minh City's Cu Chi district, with a Gmail contact address under a Vietnamese name; the developer website is silvix.app. The same developer on Play also lists Melozi (AI songwriting, 630k installs), AI Home Design (AI interior design, 740k), an IPTV player (210k), Vicado (music videos), Voicera (text-to-speech), and LumiCam (AI photo editing) (Google Play developer page and install counts, 11-06). Patrix is just one model on this production line.

How the volume was built: ramping up one tier at a time

Patrix launched 2026-08-17, and its version number is still 1.0.1,10-19 with only one update since. Almost no one installed it in the first month. Volume started in late September, and the curve since has been stepped:

(Google Play install counts and third-party estimates of daily installs, 11-06)

It stepped up tier by tier, peaking from 10-27 to 10-30, then pulled back in early November. This is the typical shape of Google App Campaigns raising and cutting budgets, not the shape of organic word of mouth.

The chart history follows the same rhythm: on 10-29 it appeared at the same time in 5 storefronts' Photography Top Free charts, ranking #27 in Nigeria, #83 in South Africa, #90 in the US, #140 in Saudi Arabia, and #177 in the UAE. The US chart only showed it for three days, from 10-29 to 10-31, and by 11-06 only Nigeria (#65) and South Africa (#141) remained (Google Play Photography Top Free chart, 10-29 to 11-06). Multiple countries rising and falling together suggests a single budget being switched on and off. The US push lasted only three days, which suggests a short test of extra spend that didn't pay back, so it was pulled.

The evidence is on its own website.

No need to guess whether it runs ads. The silvix.app homepage has a section called "Silvix Ads Reporting" that says it's an internal tool that reads Google Ads spend and conversion data from "the accounts that fund our app and game installs", "combined with our own install and revenue data, into spend, CPI, and ROAS dashboards, with alerts when campaigns drift from target", and where the account owner agrees, it can push budget changes straight back to Google Ads. The terms of service are even more direct: Silvix "builds and publishes mobile applications and games and runs paid user acquisition for them" (silvix.app homepage and terms, 11-06).

The site's own pitch is "three product lines, one publishing platform": AI apps, utility apps, and casual games, each with its own team, sharing one stack of infrastructure for publishing, data, AI inference, and paid acquisition. In plain terms: the cost of making an app has been cut down to swapping templates and package names. The real asset is the dashboard that adjusts budgets based on ROAS.

Searching "patrix" on TikTok and X returns nothing, and there's no press coverage at all. Nobody knows the name. The package name is stuffed with aigenerator, aiart, aiphoto, edit: some store-search thinking, but the 120k installs weren't searched for.

Where the money comes from

Ads only. It doesn't need to convince anyone to subscribe. As long as the ad revenue a user generates in the app exceeds what it costs to acquire them, it can keep scaling. Buying traffic in markets like Nigeria and South Africa, the cost per install is low (inferred), and the bar to crack the Photography free chart is low too, so a modest spend can get it into the top 30.

How much they actually spent, how much ad revenue each user brought in, and whether November acquisition was paid or organic — none of that is visible from outside. One detail on ratings: 1,815 ratings at 4.7 stars, a ratings-to-installs ratio of about 1.45%, which is high for a pure ad-monetized app. That suggests an in-app prompt asking for a rating. The counts at 1 stars, 2 stars and 3 stars are almost identical (57 / 38 / 38), and the reason is unclear.

Who it's competing with: launched the same day, with 5 to 15 times its volume

The "turn your selfies into a photoshoot with templates" slot has dozens of apps in the Play Photography category, most launched within the last 2026 year, and many look like they come from publishing teams in Vietnam and India (Google Play installs, 11-06):

GenFoto, which launched around the same time as Patrix, is fighting for the same Nigerian market and ranks higher. Art Studio, launched three weeks later, has 15 times Patrix's volume. Above them are established players with hundreds of millions of installs, such as Remini, ByteDance's Hypic, and Picsart, plus Google Gemini's free image editing, which popularized "AI photoshoot templates" and also brought the barrier to zero. In this red ocean, Patrix sits in the middle to lower tier.

On product, it has nothing other apps don't. Its top two complaints, that faces don't look like the user and that there are too many ads, are common to the whole category.

If it were you

One-line verdict: don't copy an AI photoshoot app. What's worth learning is Silvix's publishing method of "one ad dashboard powering a row of reskinned apps", but that path is already crowded among publishers in Vietnam and India, and Patrix itself is a middle-tier follower.

Who it's for: small teams of 1–3 people who have already run app campaigns on Google Ads, can tell CPI from ad revenue, have experience with in-app ad monetization (AdMob or a mediation layer), and are willing to treat "making an app" as assembly-line template swapping while putting their effort into acquisition and accounting.

Who it's not for: people who are only good at product and have never run ads. These apps have no organic traffic, zero discussion on TikTok or X, and without paid acquisition they get zero. Also people hoping to win on template quality: users can't tell dozens of these apps apart, and a same-period rival's volume is 5–15 times yours.

How to do it: if you do try, build the dashboard before the app. Connect Google Ads spend, installs, and ad revenue into one daily table where you can see ROAS by country and by campaign. For the app itself, use a generic image-generation API with 20–30 theme templates, doable in a week. Pick a category slot that isn't the most crowded AI photoshoot one, and look at the directions Silvix is also expanding into: AI songwriting, AI home design, AI voiceover. Choose the one with fewer peers and a lower bar on the Nigerian or South African charts.

How to acquire users: Google App Campaigns. Test with small budgets in low-cost markets like Nigeria and South Africa, and judge ROAS in a country over 7 days. If it pays back, scale up in steps, and try the US for a few days on the side. If it doesn't, shut it off and move to the next template app. Don't cram too many ad slots in: complaints about a "waterfall of commercials" directly drag conversion down.

When to get in: there's no deadline and no window. This isn't a one-time trend but a business that gets settled daily. However, each time free big-tech capabilities like Gemini upgrade, the ad revenue of template photoshoot apps gets squeezed further, so getting in later only means thinner margins.

What would mean I'm reading this wrong: (1) Patrix or GenFoto-style apps re-ramp paid acquisition after November, back to tens of thousands of installs a day, showing that low-priced ad revenue holds up and template apps' unit economics are better than assumed; (2) a new template-photo app in the photo category, without paid acquisition (organic viral TikTok videos), hits the top 50 in several stores, showing this niche still has product-level room; (3) the reverse: if apps under Mee Game Inc keep stopping updates or getting pulled, it means even teams with backing can't keep up.

---

Data and sources: Google Play store pages (Nigeria storefront, 11-06, including review text and developer info) and install counts, Photography free chart (10-29 to 11-06); daily installs are third-party estimates, spread evenly between observations; silvix.app homepage and terms of service (11-06); TikTok and X searches for "patrix" (11-06, no results). Spend, CPI, and ad revenue are not public; anything marked "inferred" or "est." in the text is an inference or estimate.