← Learn library

Houdini & Procedural Workflows / Houdini

Beginners guide to Learn 3D in Houdini 16 - SideFX 3D Video Tutorial: Procedural Interface and nodes

Channel: CGI Nerd ·

Legacy TutorialWritten guide

This tutorial is part of the CGI Nerd archive. Some software features, interfaces, or workflows may have changed since it was recorded.

Original CGI Nerd video thumbnail
Beginners guide to Learn 3D in Houdini 16 - SideFX 3D Video Tutorial: Procedural Interface and nodesCGI Nerd · Video hosted on YouTube

Playing loads YouTube content. YouTube processes data after activation.

Watch on YouTube

If the player is blocked or unavailable, use Watch on YouTube.

Software: Houdini

Learn Houdini’s interface by building a simple snowman from a small node network. The model is deliberately basic: its purpose is to make object placement, geometry operations, branching and display flags visible. By the end, one Geometry object contains several pieces, while its internal nodes retain editable construction steps.

This is a Houdini 16 introduction. The recording includes a default File node inside newly created Geometry containers and describes the then-new wire-cutting interaction. Defaults, menus and hotkeys can differ in later versions. No simulation, materials or finished character model are built here.

Create the project folders and save the scene separately

At 00:24, create a project and choose its location. The project operation establishes a directory structure for the work. It does not save the current Houdini scene: the instructor then performs a separate Save and names the scene inside that project.

Check both outcomes before modeling. There should be a project location you can find again and a saved scene file within the intended workflow. Treating folder creation as a scene save leaves the actual work unprotected.

The interface can be rearranged without changing the scene. Shelf dividers allocate room between tool groups; pane tabs can move between areas; a pane can become a Scene View, animation editor, render view, geometry spreadsheet or Python shell. Around 05:11, compare desktop layouts, save a preferred arrangement if needed, and use Reload Current Desktop to recover a familiar layout after accidental changes. This restores the desktop arrangement, not deleted scene geometry.

At 06:05, practice the navigation gestures used in the recording:

Gesture Viewport action
Space + left-button drag Orbit the perspective view
Space + middle-button drag Pan
Space + right-button drag Move the view closer or farther

The view menu changes perspective to orthographic views, while the viewport-layout menu offers single and multiple views. Keyboard shortcuts are shown too, but their effect depends on the active pane and configuration. Use the visible menu when a shortcut does not produce the expected layout.

Keep view navigation distinct from object transforms. Orbiting changes how you inspect a model; rotating an object changes the model’s placement. A useful beginner check is to navigate around an object and confirm that its transform parameters have not changed.

Read object nodes and their controls

At 09:05, use Tab in the network editor and type to find a Geometry node. In this historical setup, the new container displays a placeholder loaded by a File node inside it. The Geometry object is the container; it is not inherently the placeholder shape.

Inspect the node’s information, selectability and display controls. The display flag can hide an object even while the node and its parameters remain present. Two objects created at the same location can also appear to be one. Move one to establish whether an apparent missing object is simply overlapping another.

At 11:04, a wire between object-level nodes creates a parenting relationship. The instructor then cuts that connection with the Houdini 16 wire-cutting shortcut, Y. Later, a wire inside a Geometry object has a different role: it passes geometry between operations. The same visual idea of connecting nodes therefore needs to be understood in its network context.

Use handles for interaction and parameters for precision

At 12:04, compare the move, rotate and scale tools. The recording uses T for translation, R for rotation and E for scale, with S for selection. Hovering over a tool reveals its shortcut.

Translation arrows constrain movement to one axis; planar handles constrain it to a pair. Rotation handles offer axis-specific movement, a view-oriented outer ring and freer tumbling. Scale handles can change a single axis, a pair or all axes together. The Show Handle tool can combine controls, but its appearance depends on the selected operation.

Watch the parameter editor while using a handle. Typed numbers provide a precise alternative to dragging. In this simple unmodified example, resetting rotation values to zero restores the original orientation. Do not generalize that observation to a rig whose rest placement has been stored in pre-transforms or parent relationships.

Middle-dragging a parameter later opens the value ladder. Choose coarse increments for broad changes and finer increments for placement. The lesson demonstrates the ladder on a radius tuple so all components change together; choose the specific component when only one dimension should change.

Distinguish an object’s placement from its local geometry

At 19:02, double-click a Geometry object to enter its SOP network. Remove the placeholder File node and create a Sphere. Return to object level through the path controls, enter the other Geometry container and replace its placeholder with a Torus.

The torus appears offset because its object was moved earlier. Its internal center can still read (0, 0, 0). Around 22:08, the example object is at X = 4, while a local geometry translation of −4 can bring the visible shape back toward the scene origin. That cancellation works in the demonstrated simple transform setup; it is not a general rule for objects with additional rotation, scale or parents.

The instructor briefly changes the source primitive instead of the intended Transform SOP, then corrects the edit. Keep the two decisions separate: a primitive’s center changes its construction parameters, whereas a downstream Transform records a later geometry operation. Both may move the visible shape, but their position in the network affects subsequent work.

Move the display flag between the primitive and its Transform to inspect before and after. Selecting an earlier node does not require moving the display flag away from the final output: you can edit an upstream parameter while watching the downstream result.

At 24:07, cutting the geometry wire leaves a Transform with settings but no input geometry to operate on. Reconnect the source output to its input. This is a concrete diagnostic for an empty result: inspect data flow before assuming the transform values are wrong.

Branch one source and merge the visible pieces

At 25:03, the instructor clears the temporary objects and creates a sphere using the shelf shortcut. This creates the Geometry container and its internal source in one operation. A copied Sphere SOP inside the container is not automatically displayed alongside the first. Connect both to a Merge and put the display flag there to see the combined result.

Merge accepts multiple geometry inputs. It gathers their output; it does not, by itself, demonstrate welding their surfaces or making a watertight solid.

At 28:04, take a sphere branch through a Transform and feed that into the Merge as well. The original and transformed outputs can both contribute. The instructor then simplifies the source arrangement so multiple transformed branches derive from one sphere, producing three stacked forms. Changing the shared source affects all dependent branches; branch transforms still control their individual sizes and positions.

Trace a branch from source to Merge to understand each piece’s origin. If a sphere is duplicated unexpectedly, check whether both its original and transformed paths are feeding the Merge. If it is absent, check both the wire and the final display flag. This teaches procedural reuse without establishing packed instancing or a measured memory optimization.

Finish the primitive model and inspect both levels

At 32:00, maximize the active pane with Ctrl+B when menus are cramped. Add a Tube, connect it to the Merge, enable end caps and adjust its end radii and height into a rough carrot shape. Use its axis setting and center to point and position it on the snowman’s face. Exact dimensions are chosen visually, not supplied as a modeling specification.

At 35:11, return to object level and rename the container. Moving, rotating or scaling this one object now affects the complete snowman, while diving inside still exposes the spheres, branch transforms, tube and Merge.

The final inspection has two levels: the object transform moves the assembled model, and upstream SOP edits alter its construction. The lesson ends with that editable primitive assembly. Snow particles, richer detail and dynamic behavior are mentioned as future possibilities, not completed parts of this network.