The lamps in the furniture store had a good shape and a price tag that made no sense for a piece of molded plastic. We had a printer, we had Houdini, and we had an afternoon.
Why Houdini and not CAD
We started in Fusion 360. A revolve of a profile curve gets you a lamp shape quickly enough, but every time the silhouette needed adjusting, we were back to editing a sketch and re-running the feature tree. After three rounds of that, we switched to Houdini.
The core of the lampshade network is a revolve: a profile line revolved around a central axis to produce the outer shell. What makes it worth the setup are three ramp curves (Shape X, Shape Y, and Influence Y) that drive the silhouette at every height along the profile. Adjusting one point on a ramp updates the whole mesh immediately. Raphaela could explore shapes in real time rather than chasing a parametric history.
The groove geometry is shaped by a separate set of ramps that control groove thickness and profile. The grooves are boolean subtracted from the main form, with count exposed as a top-level parameter.
The ramp-based approach is what makes interactive shape exploration practical. Rather than rebuilding a sketch every time the proportions needed to change, Raphaela could move a handle on a curve and watch the shade update. The subnetwork exposes groove count, the groove shape ramps, cap wall thickness, and the full silhouette through a single interface.
What vase mode is, and why it suited this
Vase mode is a slicer setting that prints the entire outer shell of an object as a single continuous spiral, starting at the base and winding to the top without ever lifting the nozzle. There is no seam, no infill, and no multiple-perimeter wall. One unbroken path.
For a decorative lamp shade, that is nearly ideal. The absence of a seam line means the surface is clean all the way round. Single-wall construction keeps the shell thin enough to diffuse light, and the continuous extrusion produces a consistent surface texture with no start-stop artefacts at layer changes.
We had known about vase mode for years. After more than a decade of printing, we had somehow never had the right problem for it. A ribbed globe shell with no structural requirement was exactly that problem.
Choosing a 1 mm nozzle
The standard instinct is to reach for a 0.4 mm nozzle for anything with surface detail. A narrower bead means finer resolution, and the grooves on the shade are relatively shallow.
We went the other way. A 1 mm nozzle produces a wider bead, which means each layer line is clearly visible on the finished surface: a slight ridge at every layer boundary, a defined shadow at grazing angles. The store version had fake-printed texture. We wanted the layer lines to show, and a larger nozzle delivered that more convincingly than a finer one would have.
The practical benefits were secondary but welcome: the wider extrusion bonds more reliably at shallow overhang sections, and the overall print is faster.
Solving the flat-section problem
Vase mode has one constraint that matters here. With a single perimeter and no infill, the overhang performance depends entirely on how quickly each layer cools before the next is placed on top of it. At the equatorial sections of the shade, where the silhouette is nearly horizontal, the next layer has very little to grip.
The first print showed visible layer separation at those sections: horizontal bands where the layers had pulled apart and the gaps were visible when the lamp was lit. We adjusted the profile curve to steepen the angles at the worst points, which helped. The complete fix was print speed: slowing down at those heights gave each layer time to bond before the next pass.
Switching to Prusament opaque white for the final version also contributed. It bonds more reliably than the Anycubic white we used for early tests, and the opacity means minor surface variation disappears once the lamp is lit.
From one design to five
Project
Once the first design was dialled in (two pendants printed at the full build volume of the Prusa MK3S) the toolkit was already there. Raphaela adjusted the ramp curves to produce a rounder, more compact shape: 18 cm diameter, shorter height, fuller around the equator. We printed three of those for above the dining table.
The connector between the shade and the lamp holder is the only part not generated by the Houdini network. It is a short cylindrical adapter designed in Fusion 360 and printed separately, sized to fit a standard lamp holder from the hardware store.
Separating the connector from the shade meant we could reprint either part independently without touching the other. Both designs outperformed the store original. The opaque Prusament white diffuses light more evenly than the molded plastic, and the grooves cast fine shadow lines that shift as the eye moves around the shade.
Takeaways
Houdini is not the right tool for every fabrication job. For parts with straightforward geometry, CAD is faster and more direct. For profile-driven shapes where the design is essentially a curve being explored interactively (lamp shades, vases, organic housings, architectural profiles), a node-based approach with ramp controls is significantly faster than a parametric feature tree. We went from three rounds of sketch-editing in Fusion 360 to real-time shape exploration in Houdini. That difference compounds across iterations.
Vase mode is worth understanding properly. It works well for any closed shell where you want a seamless surface and do not need infill. The constraints to keep in mind: overhang angle (shallower than roughly 45° starts to cause problems at normal speed) and wall thickness (one perimeter means structural integrity comes from geometry, not material volume). When your shape is the right fit for vase mode, the result is cleaner than a standard multi-perimeter print.
On nozzle choice: print texture is not always a problem to solve. When the material and the process are honest parts of the design, a larger nozzle can produce a better result than trying to make something look like it was not printed at all.