We built an app where the AI invents its own generators — MakerMaker by Niixo

The core idea

Point your camera at someone. MakerMaker returns a fictional 'Corporate Operations Manual v3.2' about them — AI-written text formatted like a real document, paired with an AI-generated image. That's the first of three hand-crafted generators in the app.

The second produces a fake product spec sheet with fabricated stats. The third writes a mock breaking-news article. All three run the same pipeline: photo → Gemini 2.5 Flash for text → Pollinations.ai for image → displayed in a themed document skin.

The part we didn't plan

After shipping those three, the obvious question came up: what if the AI designed a new generator itself?

So we built a fourth mode. Tap it and the AI invents a brand-new generator on the spot — selecting genre, tone, output format, and color palette from a parameter space that works out to roughly 95 million combinations. Each run feels different from the last.

What we're honest about

About 30% of outputs are flat. Generative variance is hard to tune out entirely, especially for entertainment where the success condition is 'make someone laugh' rather than 'return the right answer.' We surface this rather than overselling.

The UI is Japanese-only. MakerMaker targets the Japanese market, where this flavor of absurdist corporate parody has an audience. Expanding to other languages isn't on the roadmap yet.

Photos aren't stored. Firebase Functions (asia-northeast1) process each request and drop the image immediately — no reason to accumulate strangers' faces for a joke app.