How to Make a Low Poly 3D Rock Generator | Houdini Tutorial

The rock generator begins with one sphere and ends with a selectable collection of irregular pieces. Its low-polygon result is produced through several changes in representation: dense geometry for displacement, volume and points for fracture, reduced geometry for manageable processing, and individual pieces for selection.

No external rock scan or finished material is used. The lesson assumes basic Houdini network navigation and works from standard geometry operations. Its central practical issue is processing cost: the first fracture attempt is slow, and the instructor changes both point count and geometry density before continuing.

Build enough surface detail to displace

The instructor creates a sphere, changes it to polygon geometry and increases its frequency. The example uses 50. He converts the sphere to a VDB, converts it back to polygons, remeshes it, and subdivides it to produce a dense surface.

This seems contrary to making a low-poly rock, but the dense stage has a different job from the final asset. It provides geometry to carry the displacement. The instructor explicitly plans to reduce the result later. Judge this stage by whether it can express the desired shape rather than by the final polygon budget.

The conversion and remeshing also alter the regularity of the sphere’s original surface pattern. Inspect the mesh after each representation change so that an unexpected result is not hidden beneath the next operation.

Let noise control the displaced position

Inside an Attribute VOP, the instructor adds Unified Noise Static, feeds position into the noise, and uses the noise result as the amount for Displace Along Normal. The displaced position feeds the geometry output’s position.

That relationship is the core of the shape generator: position determines where the noise is sampled; the noise supplies a displacement amount; the displaced positions become the new surface. The recording changes noise type and observes a more angular, rock-like form. The caption spelling of the selected noise name is uncertain, so this article does not turn it into an exact dropdown prescription.

A Multiply and Constant control the strength. With the multiplier at zero, the noise contribution disappears; raising it reveals the displacement. The instructor also varies the noise offset to obtain different shapes and frequency to change the pattern’s density.

These are distinct controls. Offset explores another part of the pattern, frequency changes its scale of variation, and multiplication changes its strength. Compare one change at a time against the silhouette, rather than assuming every increase simply makes a better rock.

The instructor briefly discusses adding normals after displacement because the shading may look odd. He postpones the final Normal node until later. A shading change and a geometry change should be inspected separately.

Scatter inside the volume and fracture the shape

The displaced rock is converted to a volume using IsoOffset, then Scatter fills that volume with points. Those points drive Voronoi Fracture, creating potential rock pieces from the larger shape.

The first attempt is expensive. The instructor reduces Scatter’s total count from 1,000 to 100 and tries again. It is still slow enough that he cancels and changes the geometry input too.

This is a useful two-part diagnosis. The number of fracture sites affects how many pieces are requested; the density of the surface affects how much geometry is being processed. Reducing only the points does not necessarily remove the cost of the dense displaced mesh.

Reduce geometry before the expensive operation

The instructor inserts Remesh after the Attribute VOP and reconnects the resulting shape to the volume/fracture workflow. The reduced geometry retains a recognizable rock silhouette, and the next fracture cooks much faster in his demonstration.

He adjusts the remesh target size and explains that a smaller target creates smaller edges and more polygons. This lets the reader frame the decision properly: preserve enough detail to carry the form while avoiding the earlier unnecessary density at the fracture stage. The video reports a qualitative speed improvement, not a measured benchmark or universal setting.

An Exploded View separates the fragments for inspection. Their exterior surfaces are denser than their newly created interiors, so he remeshes again. The second remesh serves a different purpose from the first: it makes the resulting pieces’ surface spacing more consistent after fracture.

Treat each piece as a candidate asset

The instructor adds a loop for connected pieces and discusses how fracture piece naming relates to iteration. Inside that work, Match Size uses a box reference to bring the pieces into a common placement/size context. They appear on top of one another when combined.

The captions’ references to connectivity, class and piece names are not precise enough to reconstruct every loop field. Inspect the displayed loop configuration before adapting it. The supported outcome is that the fragments are treated individually and then selected as alternative rocks.

The instructor tries a Delete node, changes to Blast, and uses Delete Non Selected to retain a chosen piece. Before building a control, inspect that selecting another piece actually changes the visible rock. This confirms that the selector is addressing the fracture pieces rather than an unrelated element group.

Expose selection and finish the preview

A Null receives an integer parameter for the rock number. In this 100-piece example the instructor sets its range to 0–99. He copies that parameter and uses Paste Relative Reference on the numeric part of the Blast selection. Changing the exposed number now selects different rocks.

The range belongs to the demonstrated piece count and naming. If those change, verify that the selector still refers to real pieces; this is not evidence that every fracture always yields an interchangeable 0–99 set.

The finishing steps add a Normal node, a brown Color, and UV Project using polar projection. The instructor inspects the UV view. He mentions later texturing, sculpting and export as possible uses, but does not demonstrate them or establish a finished production UV layout.

Inspect the generator as a sequence of decisions

Use 00:00–03:02 to revisit the dense starting surface, 03:02–07:04 for displacement controls, 07:04–10:02 for the slow fracture and geometry reduction, 10:02–14:01 for piece inspection and normalization, and 14:01 onward for selection and finishing. These ranges follow preserved captions and are reading references rather than official chapters.

The result is a set of low-poly building blocks selected from fractured, displaced geometry. Its strongest transferable idea is to use resolution where it serves a purpose, then reduce it before expensive downstream work. The exact Houdini build, ambiguous noise/loop field labels and current-version reproduction remain unverified; the article does not claim a packaged digital asset or production-ready material.