The end of Project Smile is here. I’m Dennis Karlsson, the group leader and AI programmer, and now it’s time to share my thoughts on how the project turned out. Spoiler alert: I’m happy.

The AI

The AI turned out almost exactly as I expected and wanted it to be. It can adapt to the player by learning hiding spots and get used to light, like I wanted it to. We didn’t have time to make more than two types of hiding spots, but the system of the AI supports as many types as the engine can handle.

There are some problems with the AI, I must admit. One example is that it can think that it has heard a noise when it already have investigated it. It may also investigate a sound after it has chased the player if it got interrupted during the investigation. There aren’t any bigger problems that I know of, so at least there aren’t anything that breaks the AI completely.

Player controller and interactions

The player controller turned out fairly good. It has things that was required and simply works without bugs or problems. You can sprint, crouch, light you flashlight, and interact with object to pick up items, open/close/unlock doors, light lamps, press buttons and hide. I like how we had different interaction image when looking at interactables.

The only lacking thing regarding the controller was polish. When moving, the camera doesn’t wobble at all and you don’t get much feedback while sprinting except that you move faster. Crouching could have a smooth transition instead of just switching height instantly. Hiding on spots isn’t very good either. The animation for entering a spot is very basic, and while in the spot you can’t move the camera to look around.

Regarding interactions, one thing that could’ve been better was knowing what can be interacted with. Right now, the only thing that tells that you can interact with something is the icon in the middle of the screen that pops up if you look at an interactable. We maybe should have made them pop out more graphically to show the player that this is something of interest.

The levels

The only two levels that we were able to make turned out alright. The level design doesn’t really break new grounds, but they work. The sizes of the differ greatly where the first one is rather small and the second is rather big. I like how they were openly designed, and had keys and puzzles that required the player to explore the levels which can make then find backstory pages on the way.

The biggest flaw regarding the levels was that the second one weren’t really finished regarding AI patrolling. There were path nodes all over the map, but they were used in a wrong way that could confuse the AI. It only moved in one area of the map, but if it would chase the player to another area and lose track of it, the AI would begin to use the nodes in that area. This means that the AI could get stuck in one node because this node didn’t have any information of where the AI should go next.

Things we didn’t have time for

One big thing that I was hoping to have finished in time was an ending to the story. I’ve made an ending cutscene before in another Unreal Engine 4 project, so I hoped to make another. Like I told you in my first devlog, I would be the story writer of the game. I wrote ten pages about the sad life of the psycho and how he turned mad. The ending would have the player waking up from the nightmare that the game took place in. All of the cutscene would be in first-person. There would be a voice over discussing how hard it is for the character and whether or not he would find a way to get better. The character would move towards a mirror during the monologue until you can see the face, the same face of the psycho on the nightmare without the smile. Then the character would make the same smile like in the nightmare right before the title of the game is shown with a loud and spooky noise.

We also talked about some kind of opening cutscene that would tie into a tutorial. The cutscene would feature the psycho having tied the player to a chair. A lamp in the room would then light up and scare him away, showing the player that he is afraid of light. Then there would be a scripted tutorial where the player would learn mechanics like hiding, interacting, opening doors and so on.

A big feature that we probably would have had was picking up and throwing bottles and such to attract the AI. The AI already had a system for reacting to noises so we only had to add the ability to pick up and throw for the player. The problem was that Jonathan, that was about to implement this, sadly had leave the project. The other of us didn’t have time to implement this ourselves and prioritized more important features.

Final words

All in all, I’m happy with the result we got from the project. Sure, we didn’t finish everything that we had hoped but the game turn out pretty good anyway. For eight weeks of development I think that we managed to get pretty far. I’m very happy with the AI I managed to make and the story I wrote.

A completely finished and fleshed out product of Project Smile would probably look like a mix of Outlast and Alien: Isolation, the two games that inspired this. Without any additional features and ideas it wouldn’t really be unique and stand out from those games, but I do think that it would be an amazing game anyway. That would be very interesting and fun to see, but all of us in this team has to move on to other things. This is where Project Smile ends. Thank you for reading our devlog!