For matching meshes to physics without doing it manually, add colliders to each part?
Most meshes are triangle meshes -> have whole mountain be a trinagular mesh with collisions for each triangle
Split up road and mountain, might not want to have to check the collisions with all of the scene every frame
Avoid really big single models, go for smaller models
Our physics engine might cull the collision checks to within a bounding box for us, might only have to do that manually if we're having performance issues
LOD should have different vertex counts: have both models loaded into scene, and just switch out which we're rendering based on the distance. This way, the non-visible objects never enter the rendering pipeline
EX:
Scene
/ | \
Group
|
Object/Model
| \
gltf gltf
model1 model2
model1.visible = true, model2.visible = false
- FirstPersonControls is the move for camera -- if we're going to do it manually, make cursor lock in place while still taking user mouse input. Restrict view based on relation to car's axis rather than world view
- Make 3d text a child of the car so that it's fixed in front of it
- OR, have text be rendered somewhere else in the world with a second camera with an orthographic view, could count for a second view for our advanced features.
- Look to previous assignments for how the GLSL interfaces with js
- To start on fog, look toward using a fragment shader similar to the special material in the raytracer assignment
- Have particles and shade those (individual vertices), animate it using the fragment shader (apply noise)
- Look at THREE js particle systems
TA Meeting
model1.visible = true, model2.visible = false