Objects are the top level representations of things in the scene. They have location, rotation, and scale in world space.

Models in Houdini are represented by Geometry objects at the scene level. Inside a geometry container object, geometry nodes define the geometry of the object. The geometry container object lets you translate (move, rotate, and/or scale) and shade the entire object, as opposed to individual surfaces inside the object.

When you use the shelf tools Houdini will create geometry containers with surface nodes inside for you. For example, clicking the Sphere tool on the shelf creates a new container named object_sphere1, with a sphere surface node inside named sphere1 that defines the sphere geometry.

Note that the object container and the surface nodes inside each have their own parameters. In some cases, the functionality of the parameters might seem to overlap at first. With only one sphere inside the sphere node’s Center parameter, which sets the sphere’s location, has the same effect as the container object’s Translate parameter, which moves the entire object. Similarly, the container object’s Scale parameter sets the size of the entire object, while the sphere node’s Radius parameter sets the size of the sphere inside.

Keep in mind that the normal unit of transformation (moving, rotating, and scaling) is the object. Transforming surfaces at the Geometry level (for example, using the Transform surface node) is more accurately deformation, and can potentially take much longer for Houdini to calculate than just transforming static object geometry at the scene level.

This is not to say that transforming at the geometry level is bad: often you need an object’s shape to deform over time. However, you should remember to look for opportunities to accomplish your goals by transforming objects before you look at transforming surfaces, because the former is more efficient.

Parenting makes a child object’s transformations (moves, rotations, and scales) relative to those of a parent object.

Connecting the output of object A to the input of object B makes object A the parent of object B. For example, to view the scene from the point of view of a particular object, you can parent the camera to the POV object so the camera transforms with the object.

An object which is parented to (that is, the child of) another object inherits the transformations of the parent object. Moving or rotating the parent moves/rotates the child as well.

You can also put objects inside an object subnetwork (use the tab menu to create a Subnet node at the object level). All objects inside the subnetwork inherit the transform of the subnet’s parent, if any.

Some other information is inherited from parent objects besides transformations. For example, motion blur can be inherited from a parent object. If the rendering parameters are not enabled for an object, they will be inherited from the parent object.

Each object can have only one parent, however the Blend object lets you blend between the transforms of multiple parents.

All transformations are relative to an object’s pivot point. An object scales toward/away from the pivot point and rotates around the pivot point.

The pre-transform is like an invisible parent Null inside each object. The position, scale, and rotation values in the object’s parameters are relative to the (normally hidden) values in the pre-transform.

The pre-transform can be useful, especially in character rigging, for setting a certain position, rotation, and/or scale as the "rest" or "zero" state for that object. After that, values you enter are relative to the pre-transform values.

The pre-transform uses a fixed, non-configurable transform order. If you try to copy a non-uniform scale into the pre-transform, you can get shearing.