An AI game maker for game jams, and the honest caveats

Most jam post-mortems say the same thing. Eight hours on the idea, four on setup, twenty on plumbing that nobody sees, and then a frantic hour making the game fun at the end because that was the only hour left for it.

The plumbing is the part worth attacking. Save states, menus, input handling, a pause screen, an inventory that does not corrupt itself. None of it is the game and all of it is required.

That work is exactly what Director 1 does. You describe the system and it writes the GDScript in your project. You do not read it, you play it, and when it is wrong you say what is wrong.

You build the world by hand in the builder. For a jam that is a feature rather than a cost, because roughing a space into shape with your hands is fast and iterating on it is faster.

Jam entries skew 2D, and 2D is the path a 48-hour clock likes best here. Making a project asks 2D or 3D before anything else, as two pictures, and a new 2D project is already a game: sky, a band of scenery, ground with a ledge cut into it, a platform off to one side, and a person who walks, falls and jumps under the keys you are pressing. That is hour zero rather than hour three. Widening the ground is dragging across it, and the block you drag stops a walker as soon as you let go, because the tiles, the edges and the collision all fall out of which cells are ground. Nobody spends jam hours picking corner pieces out of a palette.

Pick 3D instead and terrain, water and weather are there without setup. Scope discipline still applies either way and no tool suspends it. What changes is where the ceiling sits, because the game runs natively on the bundled Godot rather than in a tab.

One thing to settle before you plan the art. Director 1 will generate a single flat picture when you ask for one; it will not generate an animated one, so there are no frames and no sprite sheet. If your entry needs a walk cycle, that comes from the Catalogue, from an artist, or from you — and finding that out on Friday night is much cheaper than finding it out on Sunday morning.

The other jam habit that helps here is building the thing you can play first, then decorating. Get a loop that runs on hour three and you will still be able to fix it on hour forty.

You need an internet connection to build. Director 1, asset generation and world data all come over the network. If the jam is in a venue with terrible wifi, or you were planning to work on a train, that is a real constraint and you should know it now rather than at hour six.

It is a desktop app for Mac and Windows. Install it before the jam, not during. It carries its own copy of Godot, so there is nothing else to set up.

Check what your jam allows. Some jams restrict AI-assisted work or require disclosure. That is between you and the jam rules, and the code Director 1 writes lives openly in your project if you need to show it.

What you have is a Godot project. Godot exports to desktop platforms in the normal way, so the jam build is a normal build rather than something that only runs inside our app.

If your jam requires a web build specifically, be aware that this tool is not aimed at that: it is built around a game running locally with the performance that implies. Check the submission format before you commit.

The download is free and the builder is free — terrain, water, weather and lighting in 3D, painted tile ground and placed sprites in 2D, your own art either way, play and share. Indie at $10 a month unlocks the AI assistance, which in a jam is the half you lean on hardest: Director 1 writing the gameplay logic, and generating textures, sounds and props. The free plan includes a limited trial of it, so try it before the weekend rather than during.

There is no jam licence and no special tier, because it would be a fiction. It is the same app either way.

Hours one to three: decide the loop and get something moving in a rough space. Not the whole level. Enough to know what the game is.

Hours four to twenty: build the world by hand and describe the systems as you need them. Resist adding a second mechanic. The jam entries that land almost always have one idea done properly.

The last quarter: stop adding and start fixing. Sound, a title screen, whatever makes it legible to someone who was not in the room. Judges play for four minutes and the first thirty seconds decide most of it.