AI soulslike maker: commitment, recovery and the walk back

Soulslikes are among the hardest games to make well. The combat depends on animation timing at a level of precision most projects never reach, and the level design depends on a kind of spatial planning that takes real experience. Plenty of well-funded attempts have missed.

That is not a reason to avoid it. It is a reason to start with one enemy in one small area and get the exchange between the player and that enemy to feel correct before building anything else. If that exchange is not good, nothing downstream saves it.

Stage Engine helps with the part that is code, which is substantial. It does not shorten the tuning, and nothing does.

What separates this genre from other action games is that attacks cannot be cancelled. You swing, and for the next second you are committed, and so is the enemy. The whole tension comes from reading a windup and choosing what to do with the information.

That means the systems Director 1 writes for you are unusually specific: attack windows with startup, active and recovery frames, stamina that gates aggression, poise or hyper armour, invincibility frames on the dodge, lock-on that behaves sensibly when a second enemy arrives, parry timing that is strict but readable.

Each of those is a number you find by fighting. You will fight the same enemy several hundred times, which is why the code being cheap to change matters more here than in almost any other genre.

The second signature of the genre is level design that loops. A shortcut that opens back to the checkpoint, a door you saw from below an hour ago, a route that turns a punishing walk into thirty seconds. That structure is what makes the difficulty feel generous rather than cruel.

You build that by hand in the builder, placing terrain, elevation, buildings and water directly. It cannot be generated, because the entire value of the shortcut is that someone placed it precisely where relief was needed.

Plan the loop before you build the detail. Retrofitting a shortcut into a level that was not designed for one almost never works.

Stage Engine is a desktop app for Mac and Windows with Godot bundled inside. The game runs locally on your hardware, which matters in a genre where a dropped frame during an invincibility window is a lost run.

Building needs an internet connection because Director 1, asset generation and world data come over the network. Playing does not. There is no browser version.

The download is free with a free plan. Indie is $10 a month and Pro is $50, for more usage rather than a different engine.

Games in this genre are described as hard, but the ones people finish are hard in a specific way: every death is explicable. The player knows what they did, and they believe that doing it differently will work. That belief is what produces the next attempt, and the next attempt is the entire business model of the genre.

It fails when the information is missing. An attack with no readable windup, a hitbox that does not match the animation, damage that arrives from off screen. Those feel identical to a fair death in the moment and completely different afterwards, and they are what turn a hard game into an abandoned one.

So the thing to watch while testing is not whether you died. It is whether you knew why. If you cannot answer that, the fix is legibility rather than numbers, and lowering the damage will not help.