Tutorial: View bobbing: Part 2 Last edited 3 years ago2021-11-06 12:53:48 UTC

You are viewing an older revision of this wiki page. The current revision may be more detailed and up-to-date. Click here to see the current revision of this page.

Introduction

In View Bobbing: Part 1, we saw how to restore WON Half-Life weapon bobbing and the basics of custom weapon bobbing.
Here I'll show you how to bob the view itself. In other words, if we imagine the player as a camera in the world, we'll be moving this camera around, shaking it a bit, making it react to the way the player moves. Sounds good? Let's go.
You'll see the word "ref" a lot in view.cpp. If you're wondering what it means, it's "refresh". The engine's renderer is considered a "refresh module", as in, it refreshes the screen. A refdef is, thus, a refresh definition and its purpose is to define some basic parameters for rendering, such as the camera position, orientation and such. TODO: confirm this last sentence
Here's a video of the final effect with forward leaning and the inverted U weapon bobbing, so you roughly know what to expect if you followed this tutorial all the way.

The variables

The view variable is actually NOT the view I keep mentioning. It's a viewmodel, or rather, a weapon model, as you can see around line 530:
// view is the weapon model (only visible from inside body )
view = gEngfuncs.GetViewModel();
Instead, we wanna look at pparams. To be exact, pparams->vieworg is what we're looking for. It is the position of the player's view.
What basically happens at the start is this:
// refresh position
VectorCopy ( pparams->simorg, pparams->vieworg );
pparams->vieworg[2] += bobRight;
VectorAdd( pparams->vieworg, pparams->viewheight, pparams->vieworg );
The initial view origin is playerOrigin + viewHeight. We can add more vectors to this to achieve interesting effects, for example, leaning left and right, or custom recoil, or custom earthquakes. Anything you can think of that involves moving the camera (not rotating it, that'll be a different story).

It's very important to know that the view is detached from the weapon.
This means, if you suddenly go:
pparams->vieworg[2] += 40.0f;
...your view will go 40 units up, but the weapon will stay below the view.
The weapon ends up 40 units below our eyesThe weapon ends up 40 units below our eyes
No, that is not an FOV trick. The FOV was 90° in that screenshot.

It's disconnected because it does not rely on pparams->vieworg at all. It independently calculates it based on the player's origin and the view height provided by the player movement code:
// Use predicted origin as view origin.
VectorCopy ( pparams->simorg, view->origin );
view->origin[2] += ( waterOffset );
VectorAdd( view->origin, pparams->viewheight, view->origin );
So, what can we do with this? All kinds of things. Let's try leaning, Thief-style.

Leaning forward

To keep things simple, let's only let the player lean forward, and they will do that by holding Alt. I'm well aware that Alt is used for mouse strafing, but we're repurposing it for this tutorial. (in Thief, the default is Alt+W, but I wanna keep the tutorial simple enough!)

Normally, I'd assume you know the input system, but the only documentation we have is this, so I'll very briefly cover it here.

A new key for leaning

If you go to input.cpp, you'll see that there are various button variables:
...
kbutton_t    in_moveleft;
kbutton_t    in_moveright;
kbutton_t    in_strafe;
kbutton_t    in_speed;
kbutton_t    in_use;
...
We're gonna declare a new one, call it in_lean.
Then, scroll down to InitInput, and add this:
gEngfuncs.pfnAddCommand( "+lean", [] { KeyDown( &in_lean ); } );
gEngfuncs.pfnAddCommand( "-lean", [] { KeyUp( &in_lean ); } );

Adding the actual lean functionality

One way to let view.cpp know about our new key is to declare it externally:
extern kbutton_t in_lean;
Then, you will check if the key is being held, and if so, displace the view origin.
if ( in_lean.state & 1 )
{
    // Displace the view origin on a 2D plane
    // according to the player's yaw angle
    Vector offset{
        std::cosf( view->angles[YAW] * (M_PI / 180.0f) ),
        std::sinf( view->angles[YAW] * (M_PI / 180.0f) ),
        0.0f
    };

    // i will only be 0 and 1 because we're only displacing XY, not XY and Z here
    for ( i = 0; i < 2; i++ )
    {
        pparams->vieworg[i] += 32.0f * offset[i];
    }

    // Z is displaced separately
    pparams->vieworg[2] -= 4.0f;
}
This piece of code here is quite hefty, but it's really simple to understand.
The offset vector essentially represents some kind of forward direction on a 2D plane. You can think of it as pparams->forward except it's in 2D.
It then gets added to the view origin, basically displacing it. And finally, we drop the view origin a little bit lower, because if you lean forward IRL in an arc shape, you'll obviously go down a little bit.

If we tried this out in-game straight away, this would happen:
Not holding AltNot holding Alt
Holding AltHolding Alt
The weapon is gone! Well, it's not exactly gone, it's just behind us.
So, an easy way to fix this would be to modify view->origin the same way as you modified pparams->vieworg here, so like this:
if ( in_lean.state & 1 )
{
    // Displace the view origin on a 2D plane
    // according to the player's yaw angle
    Vector offset{
        std::cosf( view->angles[YAW] * (M_PI / 180.0f) ),
        std::sinf( view->angles[YAW] * (M_PI / 180.0f) ),
        0.0f
    };

    // i will only be 0 and 1 because we're only displacing XY, not XY and Z here
    for ( i = 0; i < 2; i++ )
    {
        pparams->vieworg[i] += 32.0f * offset[i];
        view->origin[i] += 32.0f * offset[i];
    }

    // Z is displaced separately
    pparams->vieworg[2] -= 4.0f;
    view->origin[2] -= 4.0f;
}
Now, there is another thing about this code, that is hard to tell from just 2 screenshots, but the transition between leaned and not leaned is VERY harsh. We need something that will smoothly float between the two states.

Interpolation in a nutshell

This was supposed to be in the next part of this tutorial series, but, I'll quickly introduce you to the concept.
Here's the thing. Say you have 2 numbers, A and B. A is 20 and B is 30. How do you find the average, middle value? Inherently, you know it's 25.
Mathematically, you'd find it this way:
x = (a + b)/2
or
x = 0.5a + 0.5b
But what if you wanted to, say, take 20% of A and 80% of B? Easy:
x = 0.2a + 0.8b
Or rather:
x = a(1-α) + b(α)
Where α is currently 0.8.

Now here's the real deal. What if you wanted to smoothly fade α from 0 to 1, thus smoothly fading the result from A to B, in a function that is executed every frame?
You can increase the value of alpha by adding the frametime to it.
alpha += frameTime;
if ( alpha > 1.0f )
   alpha = 1.0f;
Or decrease it by subtracting frametime:
alpha -= frameTime;
if ( alpha < 0.0f )
   alpha = 0.0f;
This is EXACTLY what we need to fade between "leaning" and "not leaning".

Smooth leaning

First, let's move the view displacement code outside, making it independent from the key press.
if ( in_lean.state & 1 )
{

}

{
    // Displace the view origin on a 2D plane
        // according to the player's yaw angle
    Vector offset{
        std::cosf( view->angles[YAW] * (M_PI / 180.0f) ),
        std::sinf( view->angles[YAW] * (M_PI / 180.0f) ),
        0.0f
    };

    // i will only be 0 and 1 because we're only displacing XY, not XY and Z here
    for ( i = 0; i < 2; i++ )
    {
        pparams->vieworg[i] += 32.0f * offset[i];
        view->origin[i] += 32.0f * offset[i];
    }

    // Z is displaced separately
    pparams->vieworg[2] -= 4.0f;
    view->origin[2] -= 4.0f;
}
This way, the leaning will be executed every frame, but it's good, because we can control how much of it is applied over time.
Let's declare some static variable to keep track of this transition:
static float leanFactor = 0.0f;
You can put that at the start of the function.
And then modify it, depending on the key press:
if ( in_lean.state & 1 )
{
    leanFactor += pparams->frametime;
    if ( leanFactor > 1.0f )
        leanFactor = 1.0f;
}
else
{
    leanFactor -= pparams->frametime;
    if ( leanFactor < 0.0f )
        leanFactor = 0.0f;
}
It's going to linearly interpolate towards 1 when the key is pressed, or towards 0 when the key isn't pressed. It will take it exactly 1 second. To change the speed at which this happens, you can multiply the frametime. For example, multiplying it by 4 will mean that leanFactor will go from 0 to 1 in 0.25 seconds.

Finally, we need to make use of leanFactor.
{
    // Displace the view origin on a 2D plane
        // according to the player's yaw angle
    Vector offset{
        std::cosf( view->angles[YAW] * (M_PI / 180.0f) ),
        std::sinf( view->angles[YAW] * (M_PI / 180.0f) ),
        0.0f
    };

    // i will only be 0 and 1 because we're only displacing XY, not XY and Z here
    for ( i = 0; i < 2; i++ )
    {
        pparams->vieworg[i] += 32.0f * leanFactor * offset[i];
        view->origin[i] += 32.0f * leanFactor * offset[i];
    }

    // Z is displaced separately
    pparams->vieworg[2] -= 4.0f * leanFactor;
    view->origin[2] -= 4.0f * leanFactor;
}
The rest is tweaking. For example, you might wanna use pparams->frametime * 2.5. Also, you might want to introduce some extra detail into the leaning movement:
pparams->vieworg[i] += 32.0f * leanFactor * offset[i];
view->origin[i] += 33.5f * leanFactor * offset[i];
This would make the weapon get pushed forward a little bit when leaning.
pparams->vieworg[2] -= 4.0f * leanFactor;
view->origin[2] -= 1.5f * leanFactor;
This would make the weapon go up a bit when leaning.

The U pattern

In the last tutorial, I said I'd show some formulas for different "shapes" in view bobbing. So, here is one, technically two:
x = sin( time )
y = -cos( 2 * time )
Or in code:
V_CalcBob( pparams, 1.0f, VB_SIN, bobTimes[0], bobRight, lastTimes[0] );
V_CalcBob( pparams, 2.0f, VB_COS, bobTimes[1], bobUp, lastTimes[1] );
...
for ( i = 0; i < 3; i++ )
{
    view->origin[i] += bobRight * 0.33 * pparams->right[i];
    view->origin[i] -= bobUp * 0.17 * pparams->up[i];
}
This will give you view bobbing just like in Doom.
If you want it to be more like Quake 3 and Doom 3, i.e. you want the ∩ pattern (inverted U), you just need to change:
view->origin[i] -= bobUp * 0.17 * pparams->up[i];
To:
view->origin[i] += bobUp * 0.17 * pparams->up[i];
Also, make sure you're adding bobUp to the view origin and the weapon origin. The previous tutorial applied bobRight to them, which worked, but it'd feel unnatural here.
pparams->vieworg[2] += bobUp * 0.5f;
...
view->origin.z += bobUp * 0.5f;
Notice 2 things:

In the end

Now we need to solve a number of problems: The first one is the easiest, and I'll let you figure that out. You can use linear interpolation just like we did with leaning, except we're not checking if a key is pressed, we're just gonna check if the player is on the ground with pparams->onground. I'll add this to part 3.
The second problem is out of scope for the view bobbing series, and it's a little complex, so I'll leave that for a dedicated leaning tutorial.
The third problem should actually be as simple as tweaking the frequency of bobRight and bobUp.

With minimal tweaks, you can achieve a HL2-looking weapon bob, where the weapon bobs in a U pattern. If your mod is fast-paced with lots of action, you can make it bob quite fast, but also, you can make the bob very subtle and very light if your mod is about stealth, for example.
And on the other hand, if you're making some competitive deathmatch mod, you might as well remove bobbing entirely.

In part 3, we're gonna talk about swaying, i.e. modifying view and weapon angles, which is about the same as modifying the positions with extra steps.
So until then, happy programming!

Comments

You must log in to post a comment. You can login or register a new account.