【High-Quality Live2D Tutorial for Beginners】 Eye Physics ③ (Patreon)
Content
※I apologize, but since I finished writing this article at the last minute, the translation wasn't completed in time. As a result, the images remain in Japanese and the translation hasn't been thoroughly reviewed yet.
I plan to gradually fix these issues after the new year, so please bear with me for a bit.
Also, I just learned during production that Patreon now allows embedded video posts in articles! What an amazing update! This time the video parts are GIFs, but I'll gradually switch them to embedded videos so you will be able to read them just like on Fanbox!
Hello everyone, this is Himono.
Sorry for taking a break last month! It's late, but here's the continuation of the beginner's tutorial.
Previous article is here.https://www.patreon.com/posts/141862856
This time is also a continuation of the eye physics article, but from this time onward there will be explanations of the physics calculation screen, so even if you're not rigging eye physics, it might be helpful to skim through the required sections as it could save you effort later!
-------------------------------------------------------------------------------------------------------
<Table of Contents>
[Required] About Physics Calculations
[Required] About "Type," "Effectiveness(%)" and "Reflect" in Input Settings
[Required] About "Duration," "Shaking influence," Reaction time" and "Convergence speed" in Physics Model Settings
[Optional] Rigging physics calculations for →Pupil 1
[Required] Eye Blinking and Lip Sync Settings
[Optional] Rigging physics calculations for →Pupil 2
[Optional] Rigging physics calculations for →Eyelashes
※You can easily jump to the relevant section by inputting the table of contents above with Ctrl+F (or command+F) ※Since there are many new features this time, I've explained them together at the beginning to some extent. Please return to those sections as needed when you get confused. When doing so, searching with Ctrl+F using words like "Type," "Input Settings," or "Response Speed" will make it easier to find the relevant explanations.
-----------------------------------------------------------------------------------------------------
[Required] About Physics Calculations
Since we'll actually be using the "Physics Calculations" feature from now on, I'll first explain the screen and each function. The preliminary explanation gets a bit long, so if you find it tedious, feel free to skip it for now and come back to check while rigging.
To explain again, physics calculations are a feature that automatically calculates and moves swaying of hair, clothes, etc. according to the model's movements.This time we're going to use that feature to make the pupils look teary!
To use physics calculations, select "Modeling" → "Physics Settings" from the top left menu.
Then a screen like this↓ will appear in a separate tab. You can't maximize it, but you can drag the edges of the screen to resize it, so change it to an easy-to-see size!
I've enclosed the particularly important areas in squares.Let's look at them in order.
Red square... Input Settings
This is where you set the reference parameters (input parameters) for moving with physics calculations.
I think that's a bit confusing, so let me explain it in simpler terms.For example, when do you make pupils teary? → When blinking, right? In other words, the reference parameter (input parameter) for eye physics calculations is the blinking parameter = "〇Left Eye Open/Close." With some exceptions, in my tutorials, parameters set here generally have 〇 or ◇ attached.
Other commonly used input parameters are the Angle XYZ parameters. We haven't made them yet, but these are the so-called parameters for face up/down/left/right and tilting the head. When you move your face, hair and hair accessories sway along with it, right!
Like this, you input the reference input parameters and set their influence and type (explained later) in these input settings!
Yellow square... Output Settings
This is in the same place as the input settings we just looked at. Click the "Output Settings" tab name in the upper right to switch.
As the name suggests, this is where you set the parameters that will actually be moved (output) by physics calculations. The parameters around here that we made in the previous article apply.
With some exceptions, in my tutorials, parameters set here generally have ★ or ◆ attached.
Blue square... Physics Model Settings (Pendulum Settings)
This is where you make settings to connect the input parameters entered in the red square "Input Settings" with the output parameters entered in the yellow square "Output Settings."
In the pendulum section on the right, white 〇 is the input parameter, and blue 〇 is the output parameter. You adjust various numbers in the "Pendulum Settings" section to move the pendulum.
Here's the pendulum actually moving according to the input parameter's blinking movement↓
https://drive.google.com/file/d/1u9mSyd4eFaGhsJb3Hk6807AYJMz8uo7l/view?usp=sharing
The pendulum's movement changes greatly depending on the values you set.
https://drive.google.com/file/d/1aBK06ONaSNQ06AjEXDnCH1SrvxJT3l26/view?usp=sharing
↑You can make the pendulum move violently by shortening the length and increasing the sway
https://drive.google.com/file/d/1qezMb59ZfatV-_SlpUmLb49civbOAXJ6/view?usp=sharing
↑You can make the pendulum move leisurely by lengthening the length and weakening the sway
I'll talk about the actual values in detail later.
Green square... Model Screen (※Official name unknown)
This is a screen where you can actually check the movements you've set up.
https://drive.google.com/file/d/1RmpIy1oKFacZXKRiPJqF5zrT33X_ttRW/view?usp=sharing
The physics calculation videos that I and other modelers often post are mostly capturing this screen! Besides the model's movements, you can also check how parameters are moving.
This concludes the explanation of the physics calculation screen.
[Required] About "Type," "Effectiveness(%)" and "Reflect" in Input Settings
Continuing, I'll explain the setting items in input settings.
There are three types of setting items: "Type," "Influence," and "Invert."
"Type"
Input parameters have two types: "X Position" and "Angle." What changes depending on the type is how the pendulum moves.
When you move an input parameter with type set to X Position, the pendulum moves like this:
https://drive.google.com/file/d/1uIJAGCbZb5zpWmYNcFwSOkGbFFe7p1OV/view?usp=sharing
It's like "the pendulum moves horizontally according to the input parameter. After swaying for a while, the pendulum returns to the default value."
This doesn't change even if the pendulum is long with leisurely movement.
https://drive.google.com/file/d/1lWcOl7n3h6IccSJbncy266SZRDc6EZYp/view?usp=sharing
Type "X Position" is mainly used for Angle X and Angle Y swaying (face up/down/left/right swaying).
https://drive.google.com/file/d/12WxPIHskx0iIreOTTisk20asrtv5nNgX/view?usp=sharing
↑When Angle X・Y movement stops, the swaying object's shape also returns to the default state. Since we want hair and such to return to the default state when face up/down/left/right movement stops, we use this type!
Next, when you move an input parameter with type set to Angle, the pendulum moves like this:
https://drive.google.com/file/d/17P9HRN_xVe2waQDddffs9FVi98Wev7ww/view?usp=sharing
https://drive.google.com/file/d/1-cX_2wU2q5gnTmERGJwwJVqvUL3smmIu/view?usp=sharing
It's like "the pendulum's angle tilts according to the input parameter. After swaying for a while, the pendulum stops at a value matching the angle."!
The big difference from type X Position is whether it "returns to the default state" or "stops at a value matching the angle."
Type "Angle" is mainly used for Angle Z (head tilt) swaying. When Angle Z stops while tilted, swaying objects also stop while tilted following it!
https://drive.google.com/file/d/1eA0A_eEn3ORh4hi18D9VO8Ni4vaj_vmt/view?usp=sharing
↑When Angle Z stops while tilted, you can see the swaying objects also stop while tilted following it.
In physics calculations, we create various types of swaying by skillfully using these "Position X" and "Angle" input parameter types.
"Effectiveness"(Influence)
Effectiveness allows you to decide "how much each input parameter's movement affects the output parameter." The larger the value, the greater the influence. The value is in %.
The important point is that the influence you can set for one physics calculation group has a maximum of 100% total per type!
↑In this image, there's one each of Position X and Angle input parameters, and "Position X = 〇Left Eye Open/Close" has influence 60 so there's 40 remaining, "Angle = 〇Eye Size" has influence 40 so there's 60 remaining.
For example, if you add parameters randomly here and increase the influence, you can't raise Position X above 40 or Angle above 60.
So, when there are few input parameters like this time, it's fine to set the influence to large values like 100 or 60, but as input parameters increase, you must reduce the influence accordingly and distribute it to each parameter.
↑My model's elbow physics calculations are in a terrible state.
If you thought "Won't the sway become smaller if you lower each influence value?", rest assured. By increasing the "Normalization of input" value in input settings or the " Scale " in output settings accordingly, you can cover for the reduced influence.
Remember that "Effectiveness, Normalization of input, and Scale are mutually related, and if you make one smaller the sway movement becomes smaller, but you can restore it by making another one larger" - this will save you effort later.
"Reflect"
The last setting item, Reflect, is very simple compared to the first two.As the name suggests, you can choose whether or not to invert the output parameter's movement.
↓Without Reflect check
https://drive.google.com/file/d/1YR6BLU_4wmy-VNj0kLgJw5XKLJfN7pWE/view?usp=sharing
↓With Reflect check
https://drive.google.com/file/d/1iP_0QKSvzout7QIdJWlj62TIhA8fZxYF/view?usp=sharing
As you can see, when you check Reflect, the pendulum = output parameter moves in the opposite direction to the input parameter.
By the way, there's also a place to check Reflect in the output parameters, and checking either one gives the same result. If you check both, it becomes "not" inverted.
This concludes the explanation of input settings!
[Required] About "Duration," "Shaking influence," Reaction time" and "Convergence speed" in Physics Model Settings
Next let's look at physics model settings (pendulum settings).
There are 4 types of pendulum values: "Duration," "Shaking influence," "Reaction time," and "Convergence speed."
"Duration"... The larger the value, the longer the pendulum becomes and the slower the movement.
↓Duration 3
https://drive.google.com/file/d/1n8U9fA5PxVlC2kLaIb_tzsKsvGTjIuhP/view?usp=sharing
↓Duration 20
https://drive.google.com/file/d/15OpiNpNjfmvaHR3U_EC33AKwOzb4l83N/view?usp=sharing
"Shaking influence"... The larger the value, the more it sways greatly with little movement.
↓Shaking influence 0.99
https://drive.google.com/file/d/1E6O1q_HUOMnP2pe-zisSQV-zasp0E4EC/view?usp=sharing
↓Shaking influence 0.6
https://drive.google.com/file/d/101mfLQ73FxLq674yJ2TTS72n0ipk1GhE/view?usp=sharing
"Reaction time"... The larger the value, the faster it reacts to movement.
↓Reaction time 1.5
https://drive.google.com/file/d/10Vb36fVG27figi63DDcxKi0qwycqfU_l/view?usp=sharing
↓Reaction time 0.5
https://drive.google.com/file/d/1JZ2IQW-SW3UWi589JWbvu54213eCwcLn/view?usp=sharing
"Convergence speed"... The larger the value, the faster the swaying stabilizes.
↓Convergence speed 1.5
https://drive.google.com/file/d/1Q9xqOD4N_slsme7YU34FONbUeLZ_YPNc/view?usp=sharing
↓Convergence speed 0.7
https://drive.google.com/file/d/1llpc5T25xFLG5yT1BQvfDK7CI5HGzdE-/view?usp=sharing
There's no absolute correct answer for pendulum values. I also fine-tune the values each time while watching the pendulum and actual model movements.
When you're a beginner, finding optimal values is very difficult, so it's fine to copy the values from my sample models or tutorials exactly! When doing so, keep the movement changes from the "Duration," "Shaking influence," "Reaction time," and "Convergence speed" value sizes in the corner of your mind, and try to slowly learn "what kind of movement this value produces"!
Also, if you're rigging hair swaying and such on your own first, using the default presets is convenient. Not only when creating physics calculation groups, but you can also select from the preset section of physics model settings later.
[Optional] Rigging physics calculations for →Pupil 1
The first thing we'll create is pupil physics calculations. (Sorry the first one is such a niche physics calculation... The hair, which is probably the most standard part to use physics calculations on, comes much later.)
As explained earlier, press Modeling → Physics Settings from the top left menu to open the physics calculation settings screen.
Once open, check that the Calculation FPS in the upper left is "60."If it's not, change it to 60. (As of 2025, the standard FPS for VTuber models is 60, so if possible it's better to match the physics calculations to that as well)
After checking the Calculation FPS, press the "Add" button right below it. Initially there are no physics calculations set and it's a blank state, so you add physics calculations by pressing the add button.
When you press the add button, a screen like this appears.
Make the name something easy to understand. Since we're creating pupils this time, I named it "→Pupil 1." Sometimes physics calculations are made for left and right together, and sometimes separately like this time. When making them separately, it's easy to understand later if you add notation like "←→," "Left/Right," or "LR."
The number 1 after pupil is there because I'm numbering them since we'll make 2 later.
For the input preset and physics model preset, just press OK without setting anything this time.
What these presets are is a feature that automatically inputs commonly used "input parameters like head and torso."
The physics model one is for pendulum presets. You can select fast-moving pendulums, slow-moving pendulums, multi-stage pendulums with 10 or 20 stages, etc. by default.
These presets can also be customized later. For now I've just explained them.
Now, when you press OK without setting the input preset or physics model preset, the pupil physics calculation group (※One physics calculation setting is called a "physics calculation group.") looks like this:
First, let's add input parameters from the "Add" button in input settings!
When you press the add button, a screen like this appears. Check "〇Left Eye Open/Close" and "〇Eye Size" and press the OK button.
This time it's pupil movement. We want to add a movement with a bit of recoil from blinking, and make settings to greatly change the movement with the parameter that changes eye size. So we put in these two parameters.
When you press OK it looks like this. Since the Effectiveness is still 0%, the input parameter movements aren't reflected at all yet.
We'll now change it to this!
Earlier I said "we want to add a movement with a bit of recoil from blinking." Since blinking is just "recoil," we put "〇Left Eye Open/Close" as type Position X so it returns to the default state when blinking stops. The influence is 60%.
Conversely, for "〇Eye Size," we want the pupil to stop at that position when the parameter stops. We want this kind of movement↓ So we put this as type Angle. The influence is 40, and we check Reflect.
https://drive.google.com/file/d/1bK13FCNrNkl0ya3XaZ0phj28XqmPbVQR/view?usp=sharing
Why check Reflect? Simply put, if you don't check Reflect, it moves in the opposite direction to the movement we want with 〇Eye Size.
When not checking Reflect↓
https://drive.google.com/file/d/1czx7CbW1aOltXWalWhKCrcbBD98aOwji/view?usp=sharing
Honestly, if we reversed the original blend shape movement, we wouldn't need to check Reflect. Sorry for the confusion ><
https://drive.google.com/file/d/1OS8DMpmaCFsl-dZnvajoCP0kRm-Gdm2l/view?usp=sharing
Instead, if you reverse this, you have to check Reflect for Position X. It's confusing... When you're a beginner, don't think too deeply and just use the same setting method for now.
Continuing, change the Normalization of input values in input settings like this:
The Angle value's minimum is 60 and maximum is 90, with the max and min values not symmetrical. By doing this, you can set it so the maximum value = when eyes are wide open moves (sways) more greatly.
https://drive.google.com/file/d/1xbM2y9P4KoPNcAgLrDjh_66IPRn_b6t_/view?usp=sharing
↑When you move the "〇Eye Size" parameter, if you focus on the pendulum's Angle movement, you can see that when you move 〇Eye Size to +, the Angle's movement range is larger. You can also see it's moving to the minus side because invert is checked.
https://drive.google.com/file/d/1TMN5zYUDbQR_I85dNJ-3nr6mTunJuH4u/view?usp=sharing
↑If you compare the 〇Eye Size parameter movement with the ★→Pupil 1 movement, you can see that ★→Pupil 1's - side movement range is larger than the + side movement range.
Like this, Input Normalization can not only make the movement range larger, but also create variation in the maximum and minimum value movement ranges! Let's remember this!
By the way, input parameter presets can also be selected from the preset section of input settings, not just when creating physics calculation groups.
Also, in this preset section, not only selecting presets but also registering, renaming, and deleting are possible. Blinking presets don't exist by default, so if you have time it's good to register them!
Now that input settings are done, let's set up the pendulum! Right now nothing is set so it should look like this:
There are two "Add" buttons, so press the lower add button to add a pendulum.
The default values should look like this:
Change it to this!
I changed Bounciness from 0.95→0.75, and Response Speed and Stabilization Speed from 1.0→1.5.
This creates a pendulum that "has weaker sway than the default pendulum, reacts quickly, and stabilizes immediately."
If the pupil sways too much it creates a sense of incongruity, so I always suppress the sway to about this level.
The pendulum is done with this!
Next is output settings.
Press the "Output Settings" button to switch tabs, then press the "Add" button right below to add an output parameter.
A screen similar to the input parameters earlier appears, so check "★→Pupil 1" and press OK.
Then the output parameter is set like this:
The magnification is fine as is. (In the sample data it might be a weird value like 1.045 but that's probably just an error...)
The settings for "→Pupil 1" are now complete!
Try moving "〇Left Eye Open/Close" and "〇Eye Size."
https://drive.google.com/file/d/1gK6d0DBNTEED8g97D1-48m-ndUPq7J6e/view?usp=sharing
As you can see, when you move the 〇Left Eye Open/Close parameter, the recoil makes the eyes slightly smaller the moment they open, and when you move the 〇Eye Size parameter, the eyes become much smaller.
From here let's add other physics calculations and make increasingly rich swaying!
[Required] Eye Blinking and Lip Sync Settings
Before moving to the next physics calculation, let's make one setting.
First, close the physics calculation screen by pressing the × button in the upper right.
You'll return to the original modeling screen, so click Modeling → Parameter → Blink/Lip Sync Settings from the top left menu.
A separate tab appears like this, so check Blink for the "〇Left Eye Open/Close" and "〇Right Eye Open/Close" parameters.
Scroll further down and check Lip Sync for "Mouth Deformation" and "Mouth Open/Close."
When done, press OK to complete the settings!
What does this setting do? It makes the model automatically blink with random playback and such in the physics calculation screen.
If you check Preview → Auto Blink in the top left menu of the physics calculation screen...
https://drive.google.com/file/d/1ATMKJZ-uegnyjD4vu1k8CuXaRYR3MhmS/view?usp=sharing
The model will blink like this!
Also, although we haven't made it yet, when you press random playback the mouth will also move open and closed.
[Optional] Rigging physics calculations for →Pupil 2
Next is creating Pupil 2! We could make a group from scratch again, but let's use a trick to save time.
Press the "Duplicate" button next to the group name!
Then a tab like this appears. The name has (2) added to the duplicate source.
Change this to "→Pupil 2" and press OK.
Then you can create a physics calculation group where input settings, physics model settings, and output settings are all the same as the duplicate source!
You could also save time with presets, but I often use this style of duplicating then fine-tuning.
After duplicating, change the input settings, physics model settings, and output settings like this:
Input Settings 〇Eye Size → Change type to X Position, change influence to 20%, and uncheck invert Change Angle maximum value in Input Normalization to 60 (not necessary this time since we're not using type Angle, but just in case)
Why change "〇Eye Size" from Angle to X Position? If you make this Angle, the pupil size will also change with eye size like this:
https://drive.google.com/file/d/1mVCHmvClW5EzyFPIqbMgMzUHjiyt3-Pu/view?usp=sharing
Actually, this is more correct as real human pupil movement. (Pupils change according to the brightness of light entering the eye, becoming smaller when bright = eyes wide open, and larger when dark = eyes narrowed)
However, with an anime character design like Nyanchan, changing the pupil size too much significantly changes the eye's impression, so purely based on "preference" I'm using recoil movement (=X Position) instead of size change (=Angle).
Even with Nyanchan's character design, if you prefer the size change, you can leave it as type Angle, and when creating different characters with realistic art styles, size change might suit them better! Conversely, if you think the recoil movement itself is unnecessary, you can delete the 〇Eye Size input parameter itself. (I think the distributed file had influence at 0, meaning it's in a state no different from being deleted...)
When using type Angle, keep the invert check on (if you uncheck invert, the movement will be reversed).
Physics model settings are fine as is
For output settings, change →Pupil 1 to →Pupil 2.
You can change it from the pull-down button,
Or you can delete →Pupil 1 with the delete button then add →Pupil 2 again with the add button.
When complete it looks like this! You can see the pupil moving with recoil according to the eye's movement!
https://drive.google.com/file/d/10OP3cJDkDPcEOTdnY877UDg2mFCCrDuw/view?usp=sharing
[Optional] Rigging physics calculations for →Eyelashes
Next we'll create eyelash swaying! Like before, duplicate the →Pupil 1 physics calculation group and modify it.
Input Settings 〇Eye Size → Delete 〇Left Eye Open/Close → Check invert Input Normalization → Change Angle maximum value to 60
[Explanation of why to check invert]
Normally, X Position physics calculations are used without checking invert. When you use X Position without checking invert, the movement looks like this↓
The red line is the input parameter's direction of travel. In this image it's the horizontal face movement. The blue line is the output parameter's direction of travel. In this image it's the hair. You can see the blue line moves in the opposite direction to the red line's direction of travel.
Most swaying objects like hair and clothing move the same way in reality too. The so-called law of inertia is at work.
However, for blinking we want to set the opposite of this, meaning "the red line and blue line move in the same direction."
I mentioned in this article that X Position movement returns to the default position when input parameter movement stops. In other words, for blinking, the output parameter moves like this:
The output parameter moves in the same direction as the direction of travel, then returns to the default state. A recoil movement is completed, like moving too much in the direction of travel then returning.
https://drive.google.com/file/d/17SKtNx3VrGsUS9Ow1sMXg6o4xg7F7jLU/view?usp=sharing
Not checking invert → Moves in the opposite direction to the direction of travel, creates realistic swaying
Checking invert → Moves in the same direction as the direction of travel then returns, creates cartoon-like recoil swaying
I think roughly remembering this will save you effort later!
Physics Model Settings
Let's change it dramatically. Make the pendulum 2 stages. Press the lower add button to increase pendulums, so increase to 2 stages and change to the following values:
[Explanation of values just in case] Stage 1 is the eye itself's stretching/shrinking movement, so moving too much isn't good but we want more movement than the pupil. So reduce Response Speed and Stabilization Speed from 1.5→1.4. Stage 2 is small eyelash swaying so even if you move it a lot, it doesn't create much sense of incongruity. We want to sway it more. So raise Bounciness to 0.8, and reduce Response Speed and Stabilization Speed to 1.0.
Output Settings Make this 2 stages as well.
When you add output parameters with the add button, if you check multiple parameters it automatically becomes multi-stage.
No is numbered 1, 2, 3 from top to bottom. If you want to change the order, change the ↑↓ arrow buttons and the No number!
The completed footage looks like this! The movement that was impossible in reality that we made last time becomes such a dynamic movement just by being incorporated as physics calculations. Animation is mysterious!
https://drive.google.com/file/d/17SKtNx3VrGsUS9Ow1sMXg6o4xg7F7jLU/view?usp=sharing
Unfortunately, this article ends here. Since the new feature of physics calculations appeared, there were many things to explain and I couldn't explain everything to the end ><
I tried to explain as simply as possible, but physics calculation value discussions are quite difficult for beginners in the first place, so there might be parts that become confusing while reading... If there are parts you don't understand, please feel free to ask in the comments here or on Discord!
Next time I'd like to talk about the continuation of physics calculation settings!
See you next time!