Maze TD

What it does

Classic tower defenses hand you a lane; Maze TD hands you an empty field. You build the route yourself out of cheap base towers (“Sheep”), threading enemies through mandatory checkpoints into spirals, figure-eights, and cross-map loops, then upgrade pieces of the maze in place into specialized damage dealers. There's no path rejection — wall enemies off completely and they'll attack your towers to carve through, which is where the panic and the clutch mid-wave rebuilds come from. A desktop game targeting Steam, currently a playable prototype — now in its own voxel art style: an armed flock of geese defending the meadow against predators.

Current features

• The maze is the toy — grid building with A* checkpoint routing, so the player designs the route instead of decorating a lane
• No path rejection — block freely; enemies respond by attacking towers to break through
• Build, sell, and reroute mid-wave for clutch plays
• 6 tower types (Goose, Gunner, Splash, Slow, Aura, Sniper), every one crewed by a voxel goose, vs 5 predator enemies (fox Runner, crow Swarm, turtle Armored, bear Tank, wolf Boss)
• Overlapping wave system, per-kill economy, RTS-style camera (edge pan, drag, zoom, WASD)

Tech stack

• Godot 4
• C#

What I learned

Art direction is a readability problem before it's a beauty problem. A sheep retheme died because a sheep is white fluff with a dark face — hide the fluff behind a turret and the silhouette stops reading at gameplay zoom. The goose won on silhouette alone: white body, orange bill, orange feet, legible from any angle. Voxels won the same way — grid-native, readable, and generated from code, so iterating on a model is editing text.

What's next

• Turn the spawn/checkpoint/exit squares into real assets
• Make armor types and tower tags mean something (data exists, no effect yet)
• Work the launch roadmap toward a Steam release

Build logs