forked from BilalY/Rasagar
34 lines
2.6 KiB
Markdown
34 lines
2.6 KiB
Markdown
# 3rd Person Follow
|
|
|
|
This Virtual Camera __Body__ algorithm is intended for use to implement a 3rd-person or 1st person camera. The algorithm places the camera on a mini-rig with 3 pivot points:
|
|
|
|
Pivot point 1 is the _origin_, which is the Follow target's position. When the target rotates horizontally, the rig rotates with it around this point.
|
|
|
|
Pivot point 2 is the _shoulder_, and by default is offset to one side of Pivot Point 1, to create an over-the-shoulder follow position. To make a 1st-person camera, set this offset to 0, or to whatever will give an appropriate 1st-person effect given your Follow target's position. Vertical rotations of the Follow target are transferred here, so the rig rotates horizontally about the origin, and vertically about the shoulder.
|
|
|
|
Pivot point 3 is the _hand_. This is by default offset from the shoulder, so that vertical rotations will keep the character nicely positioned on the screen. For 1st-person cameras, this can be set to 0.
|
|
|
|
Finally, the camera is positioned behind the hand, at a specifiable distance from it. The camera's rotation will always be parallel to the Follow target's rotation, but positioned behind the hand.
|
|
|
|
The camera's position and rotation are controlled by moving and rotating the Follow target, not by independent camera controls.
|
|
|
|
The 3rd-person Follow module has a built-in collision resolution system, so that if the target moves close to an obstacle, the camera position will be adjusted so that it will never be inside an obstacle.
|
|
|
|
![3rd Person Follow](images/3rdPersonFollow.png)
|
|
|
|
## Properties:
|
|
|
|
| **Property:** || **Function:** |
|
|
|:---|:---|:---|
|
|
| __Damping__ || How responsively the camera tracks the target. Each axis (camera-local) can have its own setting. Value is the approximate time it takes the camera to catch up to the target's new position. Smaller values give a more rigid effect, larger values give a squishier one |
|
|
| __Shoulder Offset__ || Position of the shoulder pivot relative to the Follow target origin. This offset is in target-local space. |
|
|
| __Vertical Arm Length__ || Vertical offset of the hand in relation to the shoulder. Arm length will affect the follow target's screen position when the camera rotates vertically. |
|
|
| __Camera Side__ || Specifies which shoulder (left, right, or in-between) the camera is on. |
|
|
| __Camera Distance__ || How far behind the hand the camera will be placed. |
|
|
| __Camera Collision Filter__ || Camera will avoid obstacles on these layers. |
|
|
| __Ignore Tag__ || Obstacles with this tag will be ignored. It is a good idea to set this field to the target's tag. |
|
|
| __Camera Radius__ || Specifies how close the camera can get to obstacles. |
|
|
|
|
|
|
|