AI JRPG maker: the systems are the easy half

More systems than almost any other genre, and most of them have to exist before a single battle can be played. Turn order. A skill list with costs and effects. Status conditions that stack and expire. Equipment that changes derived stats. A party where members join, leave and are swapped. Experience curves. Item inventory. Save points. Menus for every one of these.

That is why so many JRPG projects stall in month two. None of that is the game, all of it is required, and the person building it usually wanted to make a story about a town.

Dedicated JRPG tools solve this by shipping the systems ready-made, which is a real solution with a real cost: everything made in them plays the same, because the systems are the same and the systems are the game.

Flockbay writes the systems instead of supplying them. You describe how combat should work and Director 1 writes the GDScript into your Godot project. Nothing is chosen from a menu of supported options.

That means your combat can be genuinely odd. A turn order determined by the weapon you last used. Magic that costs health rather than a resource. A party where the fourth slot is a person who does not obey you. None of these are on any preset list, which is exactly why they are worth doing.

Then you play it and tune it. Battle systems are made of numbers and the numbers are wrong until someone sits and fixes them, which is you.

Ask anyone about a JRPG they loved and they will describe a place before they describe a battle system. The town at the start. The place where the music changed.

You build those by hand, and the app asks which kind of world before a project exists: 2D or 3D, shown as two pictures, no default, nothing created until you pick. That is the right way round in either, because a town has to be walked through to know whether it works, and no sentence you write will tell you the square is too large.

The oldest shape in the genre is the top-down one, and here it is a first choice rather than a concession. Ground is a tile layer: you drag to lay it and drag to take it back, and the tiles, the edges and the collision all fall out of which cells are ground, so a town square is walkable the moment it is painted. Whether ground stops a walker is declared per material, which is how the square and the wall behind it get painted with the same stroke and only one of them stops anybody. Top-down is not a mode you are granted, either — declare no gravity and the character body steers on both axes, because down was only ever a direction gravity gave it.

The towns get dressed from the Catalogue, which is 90.8% sprites — 55,073 objects out of 60,648 — searchable by name with a picture of each, and placed with the mouse at an exact position, turn and size. A picture sits on its own drawn bottom-centre, so an inn sits on the road instead of hovering by however much transparent margin its file carries, and draw order follows the same point, so whatever is lower down the screen is in front of what is behind it.

Pick 3D instead and you get the tools that only mean anything there — terrain from elevation, water, weather, time of day — and an overworld you can actually cross rather than something abstracted into a map screen. Same app either way, and Director 1 is told whichever world's own tools.

Branching conversations, flags that remember what you said, characters who react to events much later — all of that is code, and all of it is described rather than written.

The writing itself is yours. A tool can implement the branch. It cannot make the line land, and a JRPG lives or dies on whether you care about four people.

Desktop app for Mac and Windows, Godot bundled inside. Free to download, and the builder costs nothing. Indie at $10 a month unlocks the AI assistance — Director 1 writing the gameplay logic, plus texture, sound, prop and rig generation.

Building needs an internet connection since Director 1 and generation are online. The finished game runs locally and does not.

JRPGs are among the largest games a small team attempts, because the format implies forty hours and a continent. That expectation kills more projects in this genre than any technical problem.

The version that gets finished is usually one town, one journey out and back, and a party of three. That is a complete game and it is achievable. The continent can come after someone has played the town.

Having the systems written for you removes months of work and does not remove the content. Every map still has to be built, and every line still has to be written.