KAGIDOKO : A Deep Learning Horror Game
KAGIDOKO is an experimental horror game where every jumpscare has a reason. Collect all keys, open all doors, and get out of that house... While being stalked by a pre-trained Deep Learning Gameplay AI.
Estimated lifetime revenueLow confidence
$4.2K
$1.3Klow → high range$12.7K
Modeled estimate, not a reported figure. How we estimate →
Est. copies sold
1.5K
mid estimate
Reviews
38
Media
Screenshots & trailer
About
About the game
"There's nowhere to hide from it. All the corridors will lead to a dead end.
But it will disappear... Only if you manage to look at it under the light."
In KAGIDOKO, you are not the one who are expected to hide.
Actually, you can't.
Instead, you need to predict where the ghost could be hiding, and look right at it...
Before it sneaks up right behind you for a jumpscare.
Once the ghost decides that you can no longer be jumpscared, it will disappear.
Just to come back later, of course.
Collect all keys, open all doors, don't get jumpscared, and get out of that house.
That's all. Good luck.
The AI model (brain) in this game is roughly 3MB in size,
while YOUR brain is... About 112 million times larger than that.
Human brains are much smarter, larger, and complex than you think.
You have already outsmarted the AI by 112 million times while you are reading this.
This game is not meant to demonstrate AI supremacy or anything,
but is meant to provide continuous, unabstracted gameplay using neural networks.
In other words, I wanted a game where I literally put a real life mosquito in the game,
to make it act as the ghost for us 60 frames per second. And boom, here it is!