Okay, so after some feebacks I got from Purple Nights, I've decided to let it as it is, to be focused on the next episode.
Several things will drastically change, specially the gameplay.
Here's the list of how PN2 will be created :
-2.5D style will be preserved,
-the gameplay will be like a "building management" type,
-characters will be made in 2D-doll style,
-props & materials won't use celshading anymore,
-everything will have hand-painted textures,
-mini-games will be implemented.
At the moment, storyboard is done.
I've started the whole design of the tavern, which will be more...cute, detailled, immersive and rustic.
By the decision to remove celshading, the game will be more lightweight than PN1.
I also figured out how to implement language switch while in-game. But for now, I'm still unable to code a save system that works fine, and this will be crucial for PN2.
The base idea was to not use textures as much as possible, in order to avoid mapping.
The ground, for example, is not a simple cube, stretched and scaled.
I made sure to make the floor looks like wooden slats, with a worn & irregular aspect.
On the other hand, for paintings, I had to do base mapping, to ensure that images were displayed correctly.
But using mostly color materials for everything was not visually satisfying enough.
I had to put cel-shading to give an anime look alike style to the game. This process consists of adding an outline to every single object, characters included. But, by doing so, the size of outlined meshes is going heavier, since an outline is a duplication of all external polygons of an object, which are flipped so the visible side is reversed, giving the outlined effect.
For example, if an object have a 1Mo size, adding an outline to it will increase its size to ~ 1.7 - 1.9Mo.
Cameras & 2.5D
While coding a 2D side-scrolling camera is easy, applying the same process to a camera in a 3D environment required few... customizations.
Locking the Y rotation to ensure side-scrolling was not enough. The camera was following Yua as planned, but infortunately, since a camera's purpose is to keep the character in the center of the screen, I had to implement physical limitations, in order to limit the field of view, and consequently, making sure that the camera won't show anything outside a room.
So, I had an idea : isolate cameras in physical containers, which have been manually adjusted for every room, ensuring correct display when Yua is moving in a room.
Day cycles
This required a few lights management. Since Purple Nights has 3 cycles (day, evening & night), I had to divide all lights in 3 different folders and desactivate those which must be deactivated, while the corresponding cycle is active.
This process has also been applied to skyboxes.
As a fan of hentai art, specially hentai games (but not Visual Novels or RPGMaker games, even if I respect their creators, art style and the time they spent on their creations), I decided to give it a try.
I learned basics on Blender few years ago, modeling standard objects, for fun, to learn important notions, such as perspective, correct scaling, texture mapping, etc...
Behind the game
Modeling the game
The tavern has been made in one go. Almost every object currently displayed on game has been premade with the tavern, as an individual mesh of course, but as full part of a single tavern file.
Few objects have been added later, but at the beginning, I knew in my head exactly where each room, each object had to be placed.
All I had to do was to create it.
Here's what it looks like when opening the file :
Character creation
This was the worst part for me. I'm not skilled enough to make credible characters, so I trained hard to learn how to make good ones, in 3 different styles : Pixel 2D, "doll" 2D (with separated body parts, linked with bones to animate them), and 3D.
Unfortunately the results were very poor.
And then, one day, VRoid came out of nowhere. I tried it, and I realized that this could have been a good alternative, to compensate my biggest weakpoint.
At this point, I had to find how to convert .vrm files so that the character mesh could be displayed and modified in Blender.
Once I had the solution, all I had to do was to customize even more my VRoid creation, "Yua", in order to correct conversion & base conception defaults (such as missing seams, or polygons that aren't connected at specific points (fingers & phalanges, for example), leading to too much visible imperfections.
At first, Yua was made totally naked. I really wanted to make her clothes myself, so I learned how to create some, and I made the textures according to what I had in mind.
Here's how Yua Looks like on Blender, with visble bones of her body and clothes :
About other characters
Every other character in the game have base models owned by VRoid public creators. According to their terms of use, I choosed few ones which alteration was authorized, in order to apply specific hair, cloth colors, etc...
Nocturne is the only exception. Her base model is Yua's, but with a customized skeleton, no default face morphs, and handmade hairstyle.
Global Animation / Hentai animations
Contrary to popular belief, animation is not that hard. It takes time, of course. But once you understood how frames/keyframes are working, you've done 60% of the learning process.
From what I know, there are 2 basic animation methods in 3D : processing bone by bone, or using Inverse Kinematics (IK). Not very confortable with IKs, which slightly deforms the base mesh when not correctly configured, I've decided to make every single animation bone by bone, including Hentai ones.
what happened to 3D hentai animations ?
While handling a single character animation was easy, this were not the case for 2 characters at once.
Each one has its own skeleton and it led to coordination issues.
I chose to fuse everything : meshes and sekeltons, to make only one entity.
This was much more easy to handle, and even if I had to manage male + female animations simultaneously, results were very satisfying, very close to what I wanted to make.
Everything was going really good, and I was very confident about it. When the work was done, all I had to do was to convert my blend file + animations to FBX.
...and this is exactly HERE that troubles started.
Once the mesh has been converted, some parts of both bodies appeared heavily deformed, looking like tentacles of "Cthulhu".
"Yua and the Director's nephew" did not looked like "Yua and the Director's nephew", but more like "absurd tentacle orgy".
I checked on the internet about heavy mesh deformations after conversion, and few answers says that apparently, I forgot to do something really important before starting animations : apply position/rotation/scale of the fused meshes, making all my work, all the time spent, good for garbage.
Discouraged by this huge loss of time, I chose to try something else :
What if I try to make sprites from what I've done so far, in order to not lose my work ?
So, that's what I've done. I saved 1/2, 1/3 or 1/4 frames of each animation (depending on the total frame of each animation), then assembled all the frames on Photoshop.
This led to very few pixel gaps between some frames, but I managed to correct almost everything.
I'll tell you more about integration on Unity later. I hope I've been explicit enough. ;)
