Make a platformer with AI: getting the jump right is the whole job

Rise speed, fall speed, how long the arc holds at the top, whether the height varies with how long the button is held, whether you can still jump for a few frames after walking off a ledge, whether a jump pressed slightly early still counts when you land.

Those last two have names and almost every good platformer has them, and no amount of reading about them tells you what values are right for your game. You find them by jumping, over and over, and adjusting.

Flockbay writes all of it as GDScript in your project. You supply the verdict: too floaty, too heavy, the coyote time is too generous and it feels cheap. Each round takes a sentence and a replay.

More on this: The whole genre lives in the jump

Once the jump is right, everything else is distance. A gap the player clears with room to spare teaches nothing. A gap that needs the last frame of the jump teaches everything, and it has to sit at the exact distance, which you find by testing it.

Flockbay lays out the first rooms from the brief and plays them itself to check each gap can be cleared. Then you play, and the tuning starts: this gap one tile wider, that ledge lower. Say it and Flockbay changes it, or press Paint around the Godot editor's 2D view and drag the cells in or out yourself, with Godot's undo behind you. That loop is fast and it is the actual craft of the genre. For the last tile of a gap, your hand is often quicker than routing it through a description.

The same goes for what the player can see. Placing a hazard just off screen is a decision about the camera, not about the hazard, and you make it by looking.

More on this: Puzzle game maker: the rule is the game, and the rule gets written

Checkpoints, collectibles, moving platforms with their own timing, enemies with patrol routes, a door that needs three of something, a boss with three phases. All ordinary code, all described in a sentence.

Moving platforms are worth naming specifically because they are the classic place where a beginner gets stuck for a week — the player slides off, or gets crushed, or the platform ignores them entirely. That is a solved problem you should not be solving.

A new project starts blank. The app reads your first message and preselects 2D or 3D from it, and Flockbay asks about the game until there is a brief — for a platformer, usually one character, one jump and one short level first. Nothing is built until you approve it, and then the first twenty minutes go on jumping rather than on assembling something to jump in.

The level is a tile layer built from a coverage map: Flockbay marks which cells are ground, and the tiles, the edges and the collision all fall out of that. A block painted in a solid material stops the character the moment it exists, so changing a room is repainting cells and replaying it, whether Flockbay does it or you do with Paint. Paint the same cells in a different order and the world comes out identical, so a gap you spent an evening tuning is the same gap tomorrow.

A 3D platformer works too, and brings its own problem, which is the camera, and that is again tuning by feel. Either way the game runs locally on your own machine through the copy of Godot bundled in the app. Flockbay is a desktop app for Windows, macOS and Linux, not a browser tool.

Building needs an internet connection; playing what you built does not.

More on this: A tower defense game maker where the balance is the whole job

A flat floor, a character, and a jump. Nothing else. Make that the first thing in the brief, then spend twenty minutes doing nothing but jumping and adjusting until it is satisfying on its own, with no level around it.

That sounds like a waste of an evening and it is the single highest-value thing you can do, because every subsequent layout decision is calibrated against the jump. Change the jump later and every gap in the game is wrong.

A platformer camera has to show you where you are going without making you seasick. It should lead slightly in the direction of movement, settle rather than snap, and refuse to follow small vertical bounces.

These are all numbers, all code, and all found by playing. The tell that a camera is wrong is usually not that it looks wrong but that the game feels hard for no clear reason.

In 3D it gets worse, because now the camera has to deal with walls and corners as well. If you are starting out, this is one of the strongest arguments for making your first platformer in 2D.