Parallax Rooms with Editable Mesh

Help and Feedback Creations Feedback
1

So about a year ago I experimented with parallax rooms using surfaceGuis and imagelabels, but they never really looked that great or performed well. Now that editablemesh has advanced a lot, I’ve decided to come back and try my luck again!

You can find the imagelabels version here

Using editablemesh in theory should reduce the amount of lag cause there’s less instances and each room would be around 10 faces in total. You can see what I’ve done below.


This one uses 10 triangles and about 12 tris to make a room, so you can see the non-flat walls warping if you get too close.

This version below uses more faces/divisions for the walls so that it reduces the amount of distortion when you get too close. Though it would technically double the processing power it would still be useful for rooms that you’ll see much closer.


Here’s a video showing the higher divisions reducing the distortion more clearly

Below you can see me running a stress test, the frame rate of the windows is lowered, but in future you could probably make the frame rate adaptive so that the windows that are closer get prioritised. Each one here is getting the same frame rate, and there is no culling.

Since the editablemesh allows for me to make faces in different directions, I would be able to have one editable mesh render out up to 2000 windows each all in different positions and orientations.
If you have any suggestions for optimisations or improvements I’d love to hear about them.

5 Likes
2

I seen this used before but for reflections. But this is super cool.

2 Likes
3

Looks fantastic! Is the older open source version up to date with this one?

4

This is crazy. I love when people do stuff like this. Are the meshes actually rendered in 3d space, and then when you look behind them they’re flat? Or is it flat and your setting their normals?

I have two ideas for optimization. Have a render distance, and any vertices beyond it is only rendered every other frame. The second would be skipping every other room, cloning the mesh (same mesh content) and positioning it beside the original. So, you get two rooms for every one generated. And since they’re pretty close together, I don’t think it would affect the perspective too much.