A few years ago, I posted a blurry preview image and said, “We’re making progress on new character models, you guys!” and then never said anything else about it.
Well, at this point in time, I think I should just show you the un-blurred versions of those images. Here you go:
Before you get too attached to any of those models, let me break some very important news: after a lot of deliberation and testing, it was determined that:
- In terms of how much time has to be spent getting the models to work in the game, and how much of the current code has to be changed to support new character models, it would be much more efficient to simply upgrade the currently-existing character models rather than create entirely new character models from scratch.
- The game has had an official art style established for years now, and it would be super weird if we abruptly gave the game a completely new art style that is completely different from the pre-established art style of the game’s official artwork.
So, with those factors in mind, the decision was made to extensively modify the current models and vastly improve them, rather than completely replace the current models with new ones that feature a new art style.
There is a lot to discuss regarding the new character models, so let’s start at the beginning…
Why replace the character models in the first place?
There are four main reasons to replace Yandere Simulator’s character models:
- The body proportions of the current models are unappealing and childlike.
- The current models’ faces are unable to move very much or animate in expressive ways.
- Most of the characters have “same face” because they all use the same base model.
- It is extremely inconvenient to ever make any sort of change to the character models, since any change made to one model must also be made to every other version of that model that is wearing different clothing.
The following solutions would solve those problems:
- The character models should be adjusted to have more stylized and “adult” body proportions.
- Instead of moving bones to animate characters’ faces, we should use blendshapes.
- A character’s head and body should be two separate meshes, so that we can have a completely different head model for all characters who are significant enough to deserve unique faces.
- Clothing meshes should be separate from body meshes, so that every time we tweak an aspect of the body mesh, we don’t need to update every clothing mesh which features that body part.
What work has been done so far?
After some brief testing, it was confirmed that it’s possible and feasible to separate the body+head into two meshes and swap out heads for different characters.
As part of the process of determining and testing this, the following model was created, along with a set of 14 different head models, some of which would be used for rivals and some of which would be used by students:
(This model isn’t 100% finished or final, so if there is something about it that you don’t like, don’t freak out just yet.)
As you can probably tell, the point of this model was to get as close as possible to the body shape and art style of the game’s official character artwork, established years ago by kjech:
Here is a side-by-side comparison of the original character model and the current state of the upcoming one:
As you can see, the new character models are intended to have more stylized “anime” body proportions – slimmer waists, wider hips, thicker thighs. Also, giving each rival a completely unique head gives us the flexibility to make the rival look exactly like the concept art, instead of looking like “Ayano with different color eyes and hair.”
Take a look at this Ayano:
Feels much more like a woman now, right? Here’s a side-by-side of her original body proportions and her new ones:
The new Ayano makes the old Ayano look like a middle-schooler by comparison!
Next, check out Asu:
They actually look like the concept art now! Such a huge improvement over how they looked in the original Rival Introduction video.
Just like a chain is only as strong as its weakest link, a rival only looks as good as her hair model does. A lot of old hair models will have to be remade to live up to the new standard of quality:
A set of facial expressions have been selected to be the blendshapes that all character heads must come equipped with in order to be compatible with the game:
- Smiling Mouth
- Frowning Mouth
- Open Mouth
- Raised Eyebrows
- Sad Eyebrows
- Angry Eyebrows
- Skeptical Eyebrows
- Sleeping Closed Eyes
- Happy Closed Eyes
- Wide Open Eyes
- Optionally: Shrunken Irises. (Might want this to be controlled by a bone instead.)
With all of those different blendshapes, it should be possible to make a character perform nearly any conceivable expression.
What other problems does this solve?
Currently, we don’t have a “wearing socks” model for characters. So, when a character changes their shoes, I’m just putting a sock texture onto their shoes to make it look like they’re wearing socks. This is…pretty janky. So, how are the new character models going to solve this?
The new body mesh will have a blendshape that makes the feet transform so that all of the toes look like one solid unit instead of 5 individual toes. Here, compare the left foot to the right foot:
Slap a sock texture onto that, and nobody can tell the difference:
MUCH better than putting a sock texture onto a shoe, or a sock texture onto bare feet (making the character look like they are wearing a “glove” sock with a little pocket for each individual toe).
I know this is a kind of weird example, but it’s just one way that the act of designing new character models present us with an opportunity to solve some of the game’s oldest problems.
What other problems can this potentially solve?
Updating the game’s character models gives us the opportunity to fix poor geometry, inconsistent texel density, awkawrd UV layouts, etc. Don’t understand what that means? Here, I’ll give you an example.
Take a look at Ayano’s nose:
The line on her nose is crooked! Let’s fix it:
There, fixed! but, it’s still very pixelated, because of low texel density. So, that needs to be fixed too. In short, there’s a million little things we finally have the opportunity to fix.
For almost 9 years, Yandere Simulator has used a system where a character’s face and hair texture are included on the same texture. Obviously, this was done for the purposes of optimization (when multiple models use the same texture, the computer has to do less draw calls / less rendering passes) but the side-effect is that texture artists have to work in a small, cramped space when texturing a character’s hair. While we’re changing the characters’ models, we have the opportunity to decide “from now on, characters’ faces and characters’ hair will go on two separate textures” so that all future hair textures can be much higher in resolution.
How much can we change the characters’ proportions before problems begin to occur?
Another aspect of the game’s current character models that has bothered me for years are their necks. Anime characters are known for their long and slender necks, but the necks in Yandere Simulator are short and thick. While we’re finalizing this new character model, I’ll probably experiment with adjusting the shape of the neck and shoulders, as well…
…however, adjusting the length of the neck would wreck all of the currently-existing animations that rely on the characters’ face being in a very specific point in space. Lengthening the characters’ necks would effect…
- Animations where a character puts their head on their chin
- Animations where a character puts food into their mouth
- Animations where a character raises binoculars to their eyes
- Attack animations where a character’s neck or head is grabbed/stabbed
Etc. So, how could we change this? How could we lengthen the characters necks without changing the position of their faces? Well, I have one idea:
By drastically reducing the size of the characters’ upper trapezius, we could create the illusion that they have longer necks by effectively increasing the distance between their head and their shoulders. Get it? This is one of the things that needs to be tested before the new character models are labeled “final” and are ready to roll out.
Of course, I would also like to change other things about the characters’ bodies, like lengthening their legs and torsos to give them even more stylized anime proportions. You might think that this would probably lead to even more significant problems than lengthening the neck…but there might be a solution. I’m told that it’s easy – trivially easy – to write a script that “retargets” animations, so that all animations created for one skeleton will play flawlessly on a different skeleton that uses different proportions. I have not yet done any testing in this regard, though, so I don’t want to make any promises just yet.
For that matter, you might be wondering, “What will happen to all of the current lip-synced animations once you stop animating a character’s mouth with bones and switch to a blendshape system?” There are two potential solutions:
- Write a script that makes a characters’ mouth blendshape open and close according to the current waveform of an audio clip that is playing. (The guidance counselor is already doing this – did you notice?)
- Write a script that tells the mouth blendshape to open and close based on the presence of mouth keyframes.
However, just like the other problem mentioned above, I haven’t done any testing on this yet.
What other work needs to be done before the new character models can be implemented?
There’s somebody who I almost never mention – a certain volunteer who has contributed to Yandere Simulator in numerous very significant ways over the years:
- He created all of the motion-captured animations in the game, such as the Ayano+Kokona Befriend/Betray animation, the Senpai+Osana interaction animations, and the rival-confesses-under-tree animations.
- He created the system I use for changing a character’s model at runtime. Dozens of features (and user-made mods) would not be possible if he had never created that system for me.
- He created many of the shaders that the game uses; the shader for Osana’s tears when she is rejected, the shader for the frosted glass on the windows of the shower buildings, the shader for making shadows+blushes+bruises appear on top of a character’s face, etc.
This guy can do it all; animate, model, texture, script, make shaders, pretty much everything. He specializes in so many different things that it’s hard to figure out how to summarize him; “Technical Artist” probably comes closest. He’s probably the one volunteer who has been contributing to the game the longest, right behind Michaela Laws herself. This is one of the only guys in the world I’d trust with the extremely important job of getting the new character models perfect before they’re implemented into the game…but there’s just one problem.
Out of respect for his privacy, I’m going to be as vague as possible here, but: he was affected by the situation between Ukraine and Russia that dominated the majority of last year. Over the past year, there has been a lot of “I might be able to resume working on Yandere Sim soon!” followed by months of tense silence before I finally hear from him again. The latest news is that he’s currently in the process of moving to another country. It’s really hard to predict when he’ll actually be available to work on the new character models again, but it’s also not possible to blame him for the long wait, since it’s not his fault at all.
One potential solution is to find another technical artist who can make the final tweaks that the new character models need, but the problem is that it’s nearly impossible to find someone…
- …who cares deeply about giving Yandere Simulator the best assets it can have…
- …with his level of skill, knowledge, and expertise…
- …that I personally consider trustworthy.
…so, very little new progress has been made on the new character models over the past year.
Fortunately, I had a giant checklist of things to work on, so being unable to work on one particular thing didn’t bother me that much. But, as the checklist gets shorter and shorter, we’re getting to that point where there is very little remaining for me to focus on that is not the new character models. I still care deeply about getting the new character models into the game, but it’s an aspect of development that can’t move forward until someone adequately skilled is found to fix the problems that remain.
Still, with all of that said, I hope that this blog post has helped to assure you that a tremendous amount of work has been done on the new character models, that there is a clear plan for what remains to be done on them, and that those character models will be an objective improvement once they are finally in the game.