AI shooter generator: what you get in the first hour

Getting a controllable character with a weapon that fires and enemies that fall over is the part every tool in this space can now do, and it is the part they all show you. It is genuinely useful. It is also about two per cent of a shooter.

The remaining ninety-eight per cent is a long argument with yourself about pacing, ammo scarcity, whether the second encounter teaches the thing the fourth encounter tests, and why the assault rifle makes the pistol pointless. None of that arrives generated, and any product implying otherwise is selling you the first hour.

Stage Engine is built for the rest. The generation is there to keep the first hour short so you can get to the argument.

Code. Director 1 writes GDScript into your Godot project: controller, weapons, hit detection, enemy behaviour, objectives, UI, save state. Ask for a change in plain language and the change is written. There is no supported list of shooter features, because the layer producing features is a model writing code rather than a menu.

Assets. You can generate what fills the world, and you can bring your own models, sounds and textures because it is a normal Godot project.

What is not generated is your level layout and your judgement. Those are the two things a shooter is actually made of.

The workflow assumes you will play the game constantly. Press play, the game runs locally on the copy of Godot bundled with the app, at your machine's framerate, and you find out in ten seconds whether the change helped.

Then you say what is wrong. Too spongy. Enemies see through the wall. The reload interrupts the rhythm. That is a complete and useful specification here, and it is the kind of specification a person who does not write code can give with total confidence.

Being non-technical does not mean not knowing when a shooter feels bad. Everyone who plays shooters knows.

Stage Engine is a desktop app for Mac and Windows that you download and install. There is no browser version and no streaming. Building needs an internet connection because the model, generation and world data come over the network; playing what you built does not.

The download is free, the free plan is real, Indie is $10 a month and Pro is $50. Paid plans buy usage. Nobody gets a better engine by paying more.

Three failures come up repeatedly. The first is building ten weapons before any of them feels good, which multiplies the tuning work by ten and teaches you nothing you would not have learned from one. The second is grey boxes forever, where the levels are placeholder for so long that nobody can judge whether the game is any fun.

The third, and the most common with generated projects, is never cutting anything. Generation makes adding cheap, so people keep adding, and end up with a game that has forty features and no shape. The discipline that used to be enforced by how long things took now has to come from you.

None of these are tool problems and none of them are solved by a faster model. They are the ordinary hazards of making a shooter, and the only thing that helps is playing your own game often and being honest about it.

The most useful thing you can do with an early shooter is put it in front of one person and say nothing. Do not explain the controls, do not describe what it is going to be, and do not apologise for the placeholder art. Watch where they stop, where they look confused, and where they stop paying attention.

That session will contradict things you were certain about. It is also the fastest route to a game that works, and it costs nothing but the awkwardness of watching someone not enjoy your prototype for a few minutes.