AI game idea generator: the idea is the cheap part

Ask any general-purpose model for twenty game concepts and it will give you twenty, all plausible, several intriguing, none of them yours in any way that survives contact with a build. The reason is not that the model is bad at ideas. It is that a game idea is about four sentences long and the game is about ten thousand decisions long.

The decisions are where a game becomes distinct. How far you can fall before it hurts. Whether the door closes behind you. What the world looks like at the moment the music stops. No generator produces those, because they are not knowable until something exists to play.

So take the list. Genuinely, take it. Then pick one and find out within a day whether it is any good, which is the only test that has ever worked.

Stage Engine is a desktop app for Mac and Windows that ships with Godot inside it. It has two halves. You build the world by hand in the builder, placing terrain, water, weather, buildings and characters. Director 1 writes the GDScript for the behaviour you describe.

That split matters for testing an idea quickly. Most of what makes a concept work or fail is behavioural — the loop, the pressure, the thing that happens when you fail — and behaviour is exactly the part you can get running by describing it.

So the shape of a first session is usually: rough out a small piece of world by hand, describe the loop, play it, and find out. Often the answer is that the idea was thinner than it sounded. That is a useful answer and it took an evening rather than a month.

The concepts that hold up under building are rarely the high-concept ones. "A roguelike but underwater" collapses on contact. "You are a lighthouse keeper and the only mechanic is deciding when to sleep" survives, because it already tells you what the player does with their hands.

A good filter before you build: can you describe the thing the player is doing in the second-to-second sense, not the pitch sense? If not, the idea is still a genre, and genres do not play.

When you do have that, describing it to Director 1 is short and the result is testable fast. Vague pitches produce vague code for the same reason they produce vague games.

Stage Engine is not an idea generator and adding one would be a strange use of anyone's time. It is the thing you use after you have picked one.

It is also not a one-click generator. Nothing produces a finished game from a concept, and a page telling you otherwise is selling a demo. What you get is a much shorter distance between having an idea and knowing whether it works.

The download is free and there is a free plan, so testing an idea does not cost anything but the evening. Indie is $10 a month and Pro is $50 when you want more usage.

Director 1, asset generation and world data all come over the internet, so you cannot build on a plane. The game you have built runs locally on the bundled Godot, so once it exists you can play it anywhere.

That is worth knowing before you plan a weekend somewhere with bad wifi to finally try the idea.

Ideas change on contact with a build, and the change is usually improvement. The mechanic you thought was central turns out to be dull, and the thing you added as filler turns out to be the game. Refusing to notice that is how people spend two years building the wrong version of a good idea.

So keep the pitch loose and the build specific. The document is not the thing. The thing is what happens when someone presses a key.

Which is also why finishing something small early matters more than picking well. A concept nobody has played is a concept nobody can judge, including you.