Make your own game, and keep the your own part

The first is the template. You start from something that already works and change the parts you are allowed to change. What you end up with is recognisable as a variant of the template, and everyone who plays it can tell.

The second is one-prompt generation. You describe a game in a line, something appears, and it is a compromise between what you said and what the model guessed. You can keep describing corrections, but you are negotiating with a system that does not know what you meant, and the result converges on generic.

The third is the format. Your game only exists inside the tool that made it. That is fine until the tool changes its terms or stops existing, and then the game does too.

None of those three are necessary. They are just how most tools are built.

More on this: What one sentence can generate, and where it runs out

What gets built is decided by you before anything exists. Flockbay asks questions one round at a time, each with a recommendation you are free to turn down, and writes a brief: what gets built first, what comes after, what is not in this version. Nothing is built until you approve it. That is what keeps the result from converging on generic: the model is not guessing, it is building what you agreed.

Then it builds the world with world tools: in 3D, terrain, water, sky, weather, plants, characters and props; in 2D, tile ground, scenery behind it, and sprites at an exact position, rotation and scale. When a call is pure taste, like exactly where the house sits on the hill, the same tools are buttons around the Godot editor's own view, so you can make that one change by hand, and Godot's undo takes it back. A sentence gets you a valley; your hand can get you the exact valley.

That is more decisions than a one-prompt world asks of you, and it is meant to be. It is the reason the result is a specific place rather than a plausible one, and specific places are what people remember about games.

The judgement is also yours, all the way through. You play everything and you decide what is wrong. That never gets delegated, because nobody else knows what you were going for.

More on this: The creator in AI game creator is you

Flockbay builds the world and writes the GDScript inside your project, from the brief and your plain-language corrections, then plays its own work to check it runs. This is a genuine handover — you do not review the code, you do not read it, you do not learn a language.

Nothing is lost by giving this away, because code is not authorship. Nobody plays a game and admires the state machine. What they experience is the place and the rules, and you decided both.

A Godot project on your own disk. Godot is open source, widely used, and not going anywhere. Your project is not a proprietary save file, it is not on someone else's server, and it opens in Godot itself if you ever want it to.

Flockbay is a desktop app for Windows, macOS and Linux with Godot bundled inside it, so your game runs locally on your own hardware at your machine's framerate.

Building needs an internet connection, since the chat and asset generation come over the network. Playing what you built does not.

With the smallest version of your idea that still contains the thing you actually care about. Not a tutorial project and not a scaled-down genre exercise — the specific bit that made you want to make a game.

Put that in the brief, in this order: the place first so there is somewhere to stand, then a character moving in it, then one rule. Play it. Everything after that is the loop, and the loop is the craft.

Free to download, and every world tool is free. Flockbay is free, AI included: free hosted AI every day, or your Claude or ChatGPT plan, your own agent, or an API key. Flockbay builds from a library of ready-made objects, or with Godot's own shapes.

More on this: Make a 3D game with AI, and know exactly which part the AI did

Starting is easy and getting to eighty percent is easy. The last stretch is where projects die, and it dies for the same reason every time: the remaining work is unglamorous and the next idea is more exciting.

The counter is scope chosen honestly at the start. A game you can finish in a month is worth more than one you abandon in six, because a finished thing teaches you the whole shape of the work, including the parts nobody enjoys.

It also gives you something to show, which changes what the next project can be. Nobody plays a prototype.

More on this: Make the puzzle game you keep describing to people