4,660 installs and it charted on the US Top Grossing list: a new shell of an AI video app, and the paid-acquisition play behind it

In October 30, an AI video app called Movixel reached #49 on the Google Play US Tools top-grossing chart, with only about a thousand installs at the time. It had no website and no TikTok videos, and its rating was 2.0. It wasn't a product that took off on its own, it was another shell laid down by the same group on Play. What's worth looking at is how it climbed the charts with so few users, not the app itself.

A one-star review

In October 28, a user named "323 2jz" left a comment on Movixel's store page: "bad app. stuck on generating! want a refund asap!" 4 people marked it helpful. Scrolling down, two more: October 23, "it just doesnt work, every generation fails"; October 30, "terrible it doesn't generate like other AI app".

The Google Play listing has 11-06:11 ratings: 8 one-star, 1 four-star, 2 five-star, average 2.0. The version has been updated 1.0.1 times since October 10, and it has gone over a month without a fix.

That's the app that was still at #72 on the US Tools top-grossing chart in 11-06.

What it is

The store description is comprehensive: text-to-video, image-to-video, reference video, first-and-last frame, multi-image reference, text-to-image and image editing, templates (upload a photo of yourself and get a finished video), music and AI voiceover, outfit changes and "AI avatars", and photo packs. It's priced on credits: each generation deducts credits, and credits are bought either through auto-renewing subscriptions or one-time credit packs, listed on Play at $7.99 to $89.99 per item.

This feature list is the capability list of the 2026 video models from Kling and Seedance (inference: it's calling third-party model APIs underneath; which one is unverified). Functionally it's no different from PixVerse, Kling, Hailuo, or ByteDance's Dreamina, just worse.

The developer name on Play is "SoraBold", a near-homophone to ride on OpenAI's Sora. The legal entity is SORABOL Group Limited in Hong Kong, with a registered-agent address in an industrial building in Kwun Tong, and a phone number with a US 908 area code. The support email is a random Gmail address like [email protected]. The developer email domain demoradesign.lol opens to a templated "Hong Kong computer system design studio" that claims to have been founded 2011 years ago and to have done 480 projects, yet the company was only registered in Hong Kong in December 2023.

Where users come from: the privacy policy spells it out

Movixel leaves no trace on TikTok, X, or in the media, and it isn't on iOS either. Its privacy policy (version 2026-09-24) states quite plainly where users come from:

Facebook SDK (Meta): Used only to measure installs that come from our direct Meta (Facebook) campaigns. The SDK is disabled by default … used together with Meta's server-side Conversions API to measure conversions. Purchase events are reported server-side only.

In plain terms: the app uses Play's install referrer to determine whether the person came from a Meta ad, and only then initializes the Facebook SDK. When a user pays, the server sends a "purchase" event directly back to Meta. This is the standard setup for telling Meta to optimize ads for purchases: Meta gets the most accurate payment data, and the app ships the least SDK.

The numbers match this. Install counts on Play (googleplay library, refreshed every 2–3 days):

After launching in September 29, it got almost no volume for the first four weeks (third-party estimates show only about 1,000 installs around 10-26), then it suddenly ramped up. That's the pattern of "list it, wait for review to stabilize, then turn on spend." It has only charted on the top-grossing list, never on top free: the ads are buying paying users, not install volume. Daily installs have already dropped from roughly 1,000 at the end of October to roughly 300 in early November (estimate).

Why a few thousand installs can chart

Two things stacked up.

First, the category was chosen to be cold. Kling and PixVerse both sit in Art & Design and Photography. Movixel put itself in Tools. Tools has a low bar for top-grossing charts: at 1,094 installs (10-29) it was already #51 in the US.

Second, the ticket size is high. Credit packs start at $7.99 and go up to $89.99. A few dozen paying users, each buying one or two packs, is enough to get into the top hundred in Tools (inference; revenue is not public).

Incidentally, it also charts in the UAE, Hong Kong, the Netherlands, Poland, Singapore, and Vietnam, and nearly always at 169. That reflects Play's small-country category charts sharing one aggregate ranking, and doesn't mean it actually earns in those countries.

Who's behind it: one shell, one company, swapped when burned

Movixel's privacy policy and another app launched in Oct 17, VidAmber - AI Video Maker, are word-for-word identical, down to the device ID algorithm (ANDROID_ID plus package name, MD5) and the line "Purchase events are reported server-side only." Both policies link to api.<brand>.com/doc?docId=<seven digits>, and the document numbers climb over time (Sep 24 911211, Oct 14 1075119), which looks like one backend churning out the same documents. VidAmber's developer entity is a US engineering consultancy, "BSA Consulting Engineers, PLLC," which looks more like a borrowed or bought developer account.

Looking further back, several earlier AI video apps share the same /doc?docId= link structure. Auvrix's privacy policy has the same Meta direct-pitch clause, just worded differently (ClipNova and ReelCraft only matched on the link structure; their clauses were not compared word for word):

Inference: one team, or one white-label supplier selling to different buyers, flooding Play with "one shell, one company, one developer account". When one shell's reputation collapses or gets banned, the next one is already running. It doesn't care about retention or ratings. It's after the first credit-pack purchase.

Who it's competing with

This category is already crowded:

• Model makers' own apps: PixVerse 79.6M installs, Kling AI 34.8M (still +750k in the last 7 days), Hailuo 27.8M, ByteDance Dreamina 6.02M (googleplay library, 11-06).
• Third-party wrappers of the same kind, but two orders of magnitude bigger: Nexus - AI Video Generator with 12.4M installs, +1.85M in 7 days; Vidix 30.2M. They run heavy top-free-chart spend and monetize through ads.
• Big-tech freebies: Veo inside Gemini, OpenAI's Sora, and CapCut's AI features.

Among all these, Movixel is the smallest and weakest. It charted not by taking anyone's users, but by picking a chart nobody watches.

If it were you

The short verdict: don't copy Movixel. The shell itself has nothing worth following. What you can take away is the ad-buying and attribution playbook, applied to a product you can actually deliver.

Who it fits: anyone who already has an Android app earning from paying users and is running or about to run Meta ads.

Who it doesn't fit: indie devs who want to build "yet another AI video generator". Model makers and big tech give theirs away for free, and the wrappers on the free charts add a million or two installs a week. You have no way to win on features. It also doesn't suit anyone uncomfortable with package swapping and borrowed developer accounts. Reviews already show people demanding refunds, so this approach earns the first purchase while refunds and account bans are its cost.

How to do it: copy three things. One, use the install referrer to detect Meta-ad installs, and initialize the Facebook SDK only for those users. Send purchase events server-side through the Conversions API and bid for purchases. Two, pick categories with a low top-grossing bar (Tools-type categories), and use high-ticket one-time packs to buy chart position and store visibility. Three, list the app and let it sit, then turn on spend once it's stable.

How to acquire users: Meta optimizes for purchases, so from day one you're looking at what each paying user costs, not what each install costs.

When to move: there's no deadline. This is a general playbook, not a window.

Signs this read is wrong: ① Movixel or its sister shells keep shipping updates, ratings recover above 4 stars, and installs reach hundreds of thousands. That would mean it's not a one-shot shell but a product built in earnest, and its product selection and spend would be worth re-examining. ② A batch of AI video shells with only a few thousand installs appears on the Tools top-grossing chart. That would mean the gap in this cold chart is already known, and copying the second approach is too late. ③ Google Play takes down a whole batch of these shells at once. That would mean the shell-swapping route is closed.