The 3d that construct provides is pretty simple. Basically just some 3d objects you can place, and the ability to rotate the 3d camera. Construct’s collision system, and most behaviors are 2d so the 3d is mostly cosmetic.
You can place most object’s xy and zelevation, but you are limited to rotations around the z-axis. And the included 3dshape plugin has a limited set of different meshes you can use.
The mesh distort feature of sprites can be used to make other shapes and do rotations on other axis’s with some math. The only caveat with meshes is mesh points can only have zelevations greater than 0. It’s just something you have to workaround.
Motion, collisions and raycasts in 3d are mostly done manually with events or js. Basically done from some existing algorithm or some math.
At its lowest level you have the 3dcamera with the look at action, distort meshes and math. A lot is possible with this but it’s mostly used to do things that aren’t exactly simple.
For the math you could look up any 3d math tutorial which probably covers some vector math, rotations, spherical coordinates and some matrix math. But in practice you probably don’t need to use all of it.
Often users utilize the third party 3d object plugin to do stuff easier. It loads meshes in the gltf file format with animations. It lets you rotate objects in any way and includes behaviors for 3d physics and raycasting. It’s not free though. But it seems to try to push what you can do with 3d in construct.
In construct’s editor you cannot change the camera’s angle in 3d so it’s not ideal for designing levels in many cases. So some have experimented with the workflow of designing the level in blender.
At this time there is an issue with amd rx cards where if you have objects with transparency it will render wrong. Official response is it’s a driver bug, and amd needs to fix their driver. So you may have to deal with broken rendering for some users. Scirra doesn’t have hardware to reproduce the issue, and from what I can tell users with the issue haven’t found a way to effectively report the issue to amd to maybe get it corrected. On the plus side the author of that third party plug-in have been trying to find a patch to work and the issue.
Anyways, myst was pre-rendered as I recall. Videos transitioning as you moved around and maybe a cube view box to look around the room.
You could go all 3d though. Mostly just a matter of getting the mesh in there. To interact with stuff you could do this low tech but effective technique to do the raycasts. Basically have a 2d layer you can get the mouse from. And use a system expression to convert the xyz of the object on the 3d layer to the 2d layer and compare the distance or something.
The final bit would be the camera transitions. It should mostly be motion along a path. Same as in 2d but with z too.
Overall 3d is doable in construct but it’s mostly doing stuff yourself or using workarounds.