Camera Tool that Saves Images

Help and Feedback Scripting Support
6

Preparation

You could optionally create and use something like an octree or a BVH to store references to object(s) within your game that you would like to be captured - this would allow you to limit your world queries to a subset of the map which will reduce computation time.

This doesn’t solve having to iterate through dynamic objects, e.g. players, unless you update their reference at runtime; but it’s almost guaranteed you have significantly fewer dynamic objects than static ones.

When capturing

  1. Create or update a viewing frustum from the Camera object that the player is holding, learn more about that here

  2. Query the world using the method described in #Preparation, or alternatively, use Roblox’s spatial query API (learn more here) to find relevant game instances

  3. Using the Camera frustum, you can cull the objects from step (2) by using Frustum tests, e.g. Frustum-OBB intersection via SAT tests (see resource here)

    • I should note, however, that for your use case it probably doesn’t need to be perfect assuming you’ve added other optimisations; so you could consider a simple Frustum-Sphere test. Learn more here and here

  4. Instantiate a viewport frame such that:

    • The Viewport’s CurrentCamera instances reflects the Camera object’s transform
    • The Viewport contains the cloned game instance(s) found in step (3) - since we’re cloning and we’ve already updated the CurrentCamera coordinates, each of the instance’s coordinates should remain the same
    • This isn’t shown in the video you linked, but you can achieve a skybox-like effect in ViewportFrames if you also wanted to capture the sky in the background, e.g. example model here