Work in progress
Godot documentation is being updated to reflect the latest changes in version
4.0. Some documentation pages may
still state outdated information. This banner will tell you if you're reading one of such pages.
The contents of this page are up to date. If you can still find outdated information, please open an issue.
There are many ways to procedurally generate geometry in Godot. In this tutorial series, we will explore a few of them. Each technique has its own benefits and drawbacks, so it is best to understand each one and how it can be useful in a given situation.
All the procedural geometry generation methods described here run on the CPU. Godot doesn't support generating geometry on the GPU yet.
What is geometry?¶
Geometry is a fancy way of saying shape. In computer graphics, geometry is typically represented by an array of positions called "vertices". In Godot, geometry is represented by Meshes.
What is a Mesh?¶
Many things in Godot have mesh in their name: the Mesh, the ArrayMesh, the MeshInstance3D, the MultiMesh, and the MultiMeshInstance3D. While they are all related, they have slightly different uses.
Meshes and ArrayMeshes are resources that are drawn using a MeshInstance3D node. Resources like Meshes and ArrayMeshes cannot be added to the scene directly. A MeshInstance3D represents one instance of a mesh in your scene. You can reuse a single mesh in multiple MeshInstance3Ds to draw it in different parts of your scene with different materials or transformations (scale, rotation, position etc.).
If you are going to draw the same object many times, it can be helpful to use a MultiMesh with a MultiMeshInstance3D. MultiMeshInstance3Ds draw meshes thousands of times very cheaply by taking advantage of hardware instancing. The drawback with using a MultiMeshInstance3D is that each of your mesh's surfaces are limited to one material for all instances. It uses an instance array to store different colors and transformations for each instance, but all the instances of each surface use the same material.
What a Mesh is¶
A Mesh is composed of one or more surfaces. A surface is an array composed of multiple sub-arrays containing vertices, normals, UVs, etc. Normally the process of constructing surfaces and meshes is hidden from the user in the RenderingServer, but with ArrayMeshes, the user can construct a Mesh manually by passing in an array containing the surface information.
Each surface has its own material. Alternatively, you can override the material for all surfaces in the Mesh when you use a MeshInstance3D using the material_override property.
The surface array is an array of length
ArrayMesh.ARRAY_MAX. Each position in the array is
filled with a sub-array containing per-vertex information. For example, the array located at
ArrayMesh.ARRAY_NORMAL is a PackedVector3Array of vertex normals.
See Mesh.ArrayType for more information.
The surface array can be indexed or non-indexed. Creating a non-indexed array is as easy as not assigning
an array at the index
ArrayMesh.ARRAY_INDEX. A non-indexed array stores unique vertex information for
every triangle, meaning that when two triangles share a vertex, the vertex is duplicated in the array. An
indexed surface array only stores vertex information for each unique vertex and then also stores an array
of indices which maps out how to construct the triangles from the vertex array. In general, using an indexed
array is faster, but it means you have to share vertex data between triangles, which is not always desired
(e.g. when you want per-face normals).
Godot provides different ways of accessing and working with geometry. More information on each will be provided in the following tutorials.
The ArrayMesh resource extends Mesh to add a few different quality of life functions and, most importantly, the ability to construct a Mesh surface through scripting.
For more information about the ArrayMesh, please see the ArrayMesh tutorial.
The MeshDataTool is a resource that converts Mesh data into arrays of vertices, faces, and edges that can be modified at runtime.
For more information about the MeshDataTool, please see the MeshDataTool tutorial.
The SurfaceTool allows the creation of Meshes using an OpenGL 1.x immediate mode style interface.
For more information about the SurfaceTool, please see the SurfaceTool tutorial.
ImmediateMesh is a node that uses an immediate mode style interface (like SurfaceTool) to draw objects. The difference between ImmediateMesh and the SurfaceTool is that ImmediateMesh is a node itself that can be added to the scene tree and is drawn directly from the code, while The SurfaceTool generates a Mesh that needs to be added to a MeshInstance3D to be seen.
ImmediateMesh is useful for prototyping because of its straightforward API, but it is slow because the geometry is rebuilt every frame. It is most useful for adding simple geometry for visual debugging (e.g. by drawing lines to visualize physics raycasts etc.).
For more information about ImmediateMesh, please see the ImmediateMesh tutorial.
Which one should I use?¶
Which approach you use depends on what you are trying to do and what kind of procedure you are comfortable with.
Both SurfaceTool and ArrayMesh are excellent for generating static geometry (meshes) that don't change over time.
Using an ArrayMesh is slightly faster than using a SurfaceTool, but the API is a little more challenging.
Additionally, SurfaceTool has a few quality of life methods such as
ImmediateMesh regenerates the mesh every frame, so it is much slower than ArrayMesh or SurfaceTool. However, if you need the geometry to change every frame anyway, it provides a much easier interface that may even be a little faster than generating an ArrayMesh every frame.
The MeshDataTool is not fast, but it gives you access to all kinds of properties of the mesh that you don't get with the others (edges, faces, etc.). It is incredibly useful when you need that sort of data to transform the mesh, but it is not a good idea to use it if that extra information is not needed. The MeshDataTool is best used if you are going to be using an algorithm that requires access to the face or edge array.