An unrated app got 37k installs in a week: Framegen and its 27 "siblings"

Framegen is a small AI tool on Google Play that turns photos into short videos and also lets you try hairstyles. In three days at the end of October, its installs went from a few hundred to over 30,000, and it charted in the Top Grossing entertainment category in Thailand and Mexico. Dig deeper and it isn't really a product. It's one freshly pushed-up cell in a matrix of mass-uploaded apps rotated through paid acquisition. What's worth learning is the playbook, not the product.

What happened between Oct 30 and Nov 3

On Oct 30, Framegen: Photo to Video had 597 cumulative installs on Google Play. It launched in Sep 22 and updated to 1.0.1 in Oct 12, then never changed again. The next day the number became 14,477; on Nov 3 it was 32,119; on Nov 6 it was 37,331 (Google Play install count, 11-06). The third-party estimator AppstoreSpy puts it at roughly 5,900 installs per day from Nov 1 to 3, then down to around a thousand a day.

That same week, it peaked at #104 on Thailand's entertainment Top Grossing chart, ranked #135–#146 on Mexico's entertainment Top Grossing chart, and briefly appeared at #196 and #198 in Brazil (Google Play charts, 10-31 to 11-06).

The odd part is another number: 37k installs with zero ratings and zero reviews. The developer is "SecureVault," a personal gmail address, with no website and only this one app to its name.

What it is

The store description reads "Turn a single photo into a captivating video": upload a photo, pick an effect, make it move. The second selling point is "Try New Hairstyles" and "Explore Hair Colors." It's free to download with ads, generations cost money, and one-off in-app purchases run from $1.99 to $99.99.

Functionally it's no different from PixVerse, Kling or Hailuo, and far behind them on output quality and template library. Its privacy policy has a bolded line: "We do not use third-party AI models to process your uploaded photos; we use our own model." That can't be verified. For an app of this size, it's more likely calling someone else's API (inference).

Another mismatch: the Play listing's Data Safety section says "no data collected, no sharing with third parties," but the privacy policy says it collects device identifiers, usage data, and "attribution and install information." Attribution data is what paid campaigns use to work out which ad brought in which user.

Why it's a cell in a matrix

Following the privacy policy URL turns up the thread. Framegen's privacy page sits at file.framegenai.xyz/Framegen/privacy.html. Searching Google Play for the same URL pattern finds 27 apps, all "photo to video / AI video" apps published in 2026:

Together, these 27 apps have about 2.1M installs, and in the last ~7 days they added about 200k (Google Play install counts, 11-06). Reasons to think they're one operation, or at least built from the same template:

• The privacy page URL pattern is identical, the body text is nearly the same, and each one contains the "no third-party AI model" line.
• Each brand has its own .xyz domain, hosted at GoDaddy, with both the file. and api. subdomains pointing to AWS CloudFront. (11-06 checked Framegen, Vimo, MotoPic, Vidoza and 6 others. All the same.)
• In-app purchase tiers are nearly all $1.99–$99.99.
• Developer accounts are all personal, with names that appear to span Morocco, India, Pakistan, Bangladesh, Indonesia and Turkey, each account holding a single app. Ban one account and the others are unaffected.
• The growth curves match: a few weeks after launch, one day it jumps to several thousand installs a day, then falls back over one to two weeks. AppstoreSpy estimates Vidoza went from about 2,200 a day up to 6,300 and then down to 1,000, and MotoPic dropped from 4,600 to 1,200 a day.

From the outside you can't tell whether it's one owner or a template vendor selling to many buyers.

Where users come from

Almost certainly bought traffic (inference; no ad creatives found):

• The curve is switch-shaped: one day it's several thousand a day, a week later it's falling, and sibling apps look the same. Organic growth doesn't look like that.
• 37k installs with zero ratings. Real users, even at 0.5% rating rates, would leave a hundred or two. The whole matrix looks like this. The largest, Vimo AI, has 346k installs and only 784 ratings.
• The privacy policy mentions attribution data; the sibling apps' privacy pages name Meta's SDK and Singular, an attribution company used in paid acquisition.
• You can't find it on social media: searching "framegen" on TikTok and X returns nothing relevant, and there's no press coverage.

Store search brings in some volume too. Every app name is "brand name + Photo to Video / AI Video Maker." Right after launch, they showed up at #130–#190 on the free entertainment charts in Italy, Vietnam and Thailand.

Where the money comes from

Free to download, ad-supported, with video generation sold as credits or subscriptions. From $1.99 per item up to $99.99. Revenue isn't public.

What's estimable is the entry barrier: in Thailand and Mexico's entertainment Top Grossing charts, 100 people don't need much money. Dozens of installs a day can get one charted (estimated). Apps in this matrix have charted in 24 countries and regions, with their best finish at #29–38 on category Top Grossing charts (Brazil 29, Nigeria 33, Indonesia 33, Mexico 37). They target markets like Southeast Asia and Latin America, where paid acquisition is cheap and chart barriers are low. They skip the US store.

The math of the operation looks like this: when an old cell's installs fall off, its ratings slide, or its account gets banned, they swap in a new account, new domain and new name, push another cell up, and run another wave. The October batch of new cells includes Framegen (9-22), PortraLabs (9-23), FantasyPic (10-10), MotoPic and Clipsy (10-13), and Video Me (10-21). Counting from Vimo AI's launch in February, the playbook has run for 9 months.

Who it's competing with

Photo-to-video is one of the hottest AI entertainment features this year. The demand is real, and so is the crowding (Google Play install counts, 11-06):

Framegen doesn't rank on this table. It gets its volume from distribution, not the product.

If it were you

C: There's a playbook to learn, but a solo build of an app like this won't keep up. Framegen has nothing of its own worth copying. What's valuable is the pipeline behind it: one codebase, dozens of shells, dozens of developer accounts, and paid acquisition rotated through cheap countries.

Who it suits: small teams that have already shipped utility apps on Google Play, run Google App Campaigns or Meta app ads, and can work out what each install costs. You should have a video generation API in hand and know what each video costs you.

Who it doesn't suit: anyone aiming to build a "better photo-to-video app" (PixVerse, Kling, Hailuo and big-tech free features already dominate, and you can't out-product them); anyone without ad experience who's counting on organic store traffic (the category has none, and Framegen's 37k installs with zero ratings prove it); and anyone who can't handle accounts being banned or apps being pulled.

How to do it: copy the structure, not this one app. Use one codebase, change name, icon and headline feature (photo-to-video, hairstyle try-on, old-photo restoration, pets that move) to make several shells. Give each shell its own domain and privacy page. Work out the generation cost first: video API calls cost cents each, and your first-purchase price has to cover that. Keep your Data Safety declarations and privacy policy consistent. Framegen's two disagree, and that's a ready-made reason for Google to pull it.

How to get users: pick one country with a low Top Grossing threshold (Thailand, Mexico, Indonesia, Brazil) and spend a small budget for a week or two. See whether you can chart in the category Top Grossing and whether in-app purchases recover your cost per install. If it works, scale up. If not, stop and move to the next shell.

How early you need to be: there's no hard deadline, but the window is narrowing. Kling, Hailuo and PixVerse are all spending heavily on ads, and Meta and Google are moving the same features to free. The later you start, the harder it gets to buy users who will pay.

What would show you've got it wrong: (1) installs for this October batch of new cells (Framegen, MotoPic, FantasyPic) stop growing a month later, and no new cells launch, meaning the playbook has outrun the matrix's own capacity; (2) Google pulls matrix apps in bulk, meaning the swap-the-skin route is blocked; (3) the reverse: if 2027 early in the year new cells were still launching monthly and charting, that means unit economics are positive and worth working through carefully.

Sources

• Google Play pages, install counts, charts: 2026-10-29 to 11-06. Daily install estimates from AppstoreSpy.
• Privacy policies: file.framegenai.xyz/Framegen/privacy.html (effective 8-31), plus similar pages for Vimo AI and MotoPic.
• DNS lookups: 11-06.
• TikTok, X, web search: 11-06. No Framegen-related content found.
• Wherever "inference" or "estimate" appears, the claim is an inference or estimate. No public data on revenue, retention, or which model is used.