Blender Geometry Nodes: Instance Objects on Points and Realize Instances

What you’ll learn

This guide explains how Instance on Points and Realize Instances divide the work in Blender Geometry Nodes. The first node places a geometry as an instance on each existing point, using a reference to it instead of duplicating the same data over and over. This makes it a useful way to add many similar elements to a scene.

The second node converts these instances into real geometry data. This makes it possible to modify instances individually, but can increase processing costs, especially when complex geometry is instanced many times. In architectural visualization, understanding the choice between these two stages helps you plan how to manage repeated elements.

Requirements

• You’ll need Blender and Geometry Nodes. The source documents are from the current version of Blender’s online manual; as they don’t specify compatibility with a particular Blender version, check the manual for the version you’re using as well.

• You’ll need geometry that contains a Point domain. According to the manual, this can be a mesh, point cloud, or curve control points.

• Prepare the geometry to place on the points. It can be real geometry or geometry that contains other instances.

• The sources don’t specify any fees, account requirements, or special hardware requirements. Don’t assume any additional conditions for these.

This walkthrough describes a general node setup using existing point geometry and the geometry to be placed. Rather than specifying a particular interface menu path or version-specific operation name, it focuses on the documented inputs and outputs of the two nodes.

Step by step: placing instances on points

1. Choose the point geometry. Connect geometry with a Point domain to the Points input of the Instance on Points node. The points’ positions affect the transforms of the resulting instances. Meshes, point clouds, and curve control points are among the geometry types this node can work with.

2. Choose the geometry to instance. Connect the geometry you want to place to the Instance input. The node adds an instance of this geometry to each selected point. When the same geometry is used on many points, instancing references the underlying geometry data rather than copying it.

3. Control which points are used. The Selection input determines whether an instance is created at each point. True values produce instances; false values skip the corresponding point. This lets you work with the points specified by the selection data without having to use every point.

4. Set the rotation and size inputs. Rotation determines each instance’s Euler rotation, while Scale determines its size. The Rotation input can be connected to the rotation output of nodes such as Distribute Points on Faces or Curve to Points. Align Rotation to Vector can be used to create an Euler rotation from a direction vector. These are the connection options described in the source; it doesn’t prescribe a single correct value for a particular architectural layout.

5. If needed, choose from a list of instances. The Instance input can contain multiple instances rather than a single geometry. When Pick Instances is enabled, the node places the instance specified by Instance Index at each point instead of the entire input. The manual says this option is designed to be used with the Collection Info node. Negative Instance Index values, or values outside the list’s range, wrap around to the other end of the list.

6. Account for nested structures if the point geometry already contains instances. If the input geometry contains instances, the node can also add new instances to the points inside them. The new instances receive the transform from the Rotation and Scale inputs, and are also affected by the transforms of their parent instances. If the result looks unexpected, check for this possibility of nested instancing.

The Instance on Points output is of type Instances. At this stage, you work with instances that reference the same underlying geometry rather than modifying each one individually. Attributes on the point geometry can also be used on the instance domain of the created instances.

Converting instances into real geometry

1. Decide whether conversion is necessary. Add a Realize Instances node to the node tree if the instances need to be affected individually. It converts the instances in the input into real geometry data, making it possible to perform operations on them that differ from changes applied uniformly to the same geometry before conversion.

2. Choose which instance levels to convert. Selection determines which top-level instances are converted into real geometry. Realize All converts every nested instance level within each top-level instance and overrides the Depth input. If Realize All isn’t enabled, Depth determines how many nested levels are converted for each top-level instance.

3. Consider attributes in the output. For attributes carried over from instances to geometry, the most complex data type is chosen for the output when multiple inputs are combined. If an attribute exists both on the original geometry and on an instance, the original geometry’s values take priority. The id attribute is also handled specially to avoid duplicate values. According to the source manual, vertex groups are also preserved when realizing instances or joining geometries.

Common mistakes

• Realizing every instance straight away: This makes it possible to edit instances individually, but can significantly hurt performance when there are many instances of complex geometry. Keep instances as instances at first, and identify the stage where conversion is actually needed.

• Overlooking nested instances: If the point input already contains instances, new instances may also be created at lower levels. Parent instance transforms can affect the result, too.

• Assuming Realize All and Depth both take effect: When Realize All is enabled, the Depth value is ignored. Keep this in mind if you want to use Depth to limit the number of nested levels converted.

• Assuming where attributes come from: If the original geometry and an instance have attributes with the same name, the original geometry’s values take priority. If data types differ across inputs, the more complex type is selected for the output.

• Misunderstanding Pick Instances: When this option is disabled, the geometry in the Instance input is added to every selected point. When enabled, instances are selected from the list using Instance Index.

Next steps

In your next test, create two separate node setups using the same point geometry and instance geometry: leave the Instance on Points output as instances in one, and convert it to real geometry with Realize Instances in the other. In both setups, observe the point selection, Rotation, and Scale inputs separately. Then add the conversion step only when instances need to be edited individually.

If you’re working with instance lists, explore the relationship between Pick Instances, Instance Index, and Collection Info as well. For nested instancing, compare the different behaviors of Realize All and Depth in a small test. Keep in mind that performance can drop when working with complex geometry and many instances; the source documents don’t provide a specific threshold or performance measurement.

This approach is suitable for repeating the same geometry across many points in architectural, interior design, and archviz workflows. Decide whether to convert instances into real geometry based on whether later operations need to modify each element individually. Checking the documentation for the Blender version you’re using also helps ensure node behavior matches that version.

Source and license

This guide is adapted into Turkish from the descriptions of Instance on Points and Realize Instances in Blender’s official Geometry Nodes documentation. Both sources are published under the CC BY-SA 4.0 license.