Making a twin stick shooter without writing the controller

Nobody finishes a twin stick shooter because of the story. They finish it because moving is pleasant, dodging is readable, and the gun makes a satisfying thing happen. That is acceleration curves, dead zones, recoil, hit pause, knockback, screen shake and about a dozen other values that are individually trivial and collectively the entire experience.

Those values are exactly what a person who does not code normally cannot touch. They live inside a script, and every adjustment is a round trip through someone who can edit it. Flockbay collapses that: you say the dash is too floaty and ends too far away, and Flockbay changes the code that governs it.

You are the one who decides when it is right. Not writing the controller does not mean not having an opinion about the controller. It means the opinion is what you contribute, and it is the part that actually needs a human.

Twin stick arenas are small and load-bearing. Cover placement decides whether the room is tense or unfair. A pillar in the middle changes the dominant strategy. Sightlines determine whether an enemy that spawns behind you is a threat or a cheap shot.

So you place those yourself, directly, in the builder: walls, props and lighting on terrain in a 3D arena, painted tile ground and dragged sprites in a flat one. You are not describing a room and hoping. You are moving the pillar two metres and looking at it from the player camera.

You can build a single arena and iterate on it forever, or lay out a run of connected rooms if you are heading somewhere roguelike. The tool does not have an opinion about which, because there is no template underneath prescribing a shape.

Ask for an enemy that charges in a straight line and overshoots, and one that circles at a fixed radius and fires in bursts, and one that splits into two smaller versions when killed. Flockbay writes the GDScript for each inside your project.

Weapons work the same way. A shotgun with a tight cone and heavy knockback. A beam that ramps damage while held and overheats. A weapon that fires backwards and is meant to be used for movement. Describe the behaviour, play it, say what is wrong.

The revision loop is the normal way to work here rather than a sign something failed. Most of these behaviours are wrong on the first pass because the right numbers are not knowable in advance. They are found by playing.

Flockbay is a desktop app for Windows and ships with Godot inside it. Press play and the game runs locally at your machine's framerate. This is not a small detail for a genre where two hundred bullets on screen is a normal Tuesday and where input latency is the difference between tight and broken.

A browser tool cannot honestly promise you that, which is why browser tools tend to steer you toward calmer scenes. Running locally is what makes bullet density a design choice rather than a budget.

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

A twin stick shooter with no progression is a technical demo. What keeps somebody in it is the sense that the next run has something in it: a weapon they have not held, a room shape they have not learned, a modifier that changes what a good build looks like.

That is systems work rather than art, and it is the sort of thing that spirals if you have to write it yourself. Unlock conditions, persistent currency between runs, a pool of modifiers that combine without producing an unplayable mess, a seed so a good run can be shared. Each is a request in plain language and Flockbay writes the GDScript for it into your project.

The part you keep is deciding which of them matter. A pool of forty modifiers where thirty are boring is worse than eight that all change your decisions, and no tool can tell you which eight they are. You find that by playing your own game more than is comfortable.