TL;DR
Get home appliances delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A maker is using Google’s Gemini image-generation and coding tools to help design components for a Fallout-inspired Halloween prop with his son. Gemini produced editable OpenSCAD code for some parts, but also missed design instructions, and the project remains unfinished as Halloween approaches.
A maker is using Google Gemini to help design 3D-printable parts for a Fallout-inspired Halloween prop he is building with his son, according to a report published by ToolGuyd. The project shows a practical, limited use of generative AI: producing editable OpenSCAD code and helping visualize shapes, while leaving the maker to correct errors, choose materials and assemble the final object.
The project began as a costume plan tied to the son’s interest in Fallout 4. The maker ruled out building full power armor within the available time and instead settled on a character look with an Enclave-decorated chest piece and a handmade prop. A planned motorcycle chest protector ordered online was delayed and appeared to be lost, so the costume’s armor elements may instead use foam and craft materials.
For the prop, the maker asked Gemini for help with a truncated cone and mounting ring. After several iterations, the generated OpenSCAD code represented the shape as two connected truncated cones. Because the code uses adjustable parameters, the maker could change dimensions and geometry without starting the model from scratch. A test print was described as good enough to guide the next revision, although the filament used was wet and was being dried.
The maker also used AI-generated code to develop a handle and other details, then edited parameters directly, including adding mounting-plate holes and changing slots. The planned assembly may use a small wooden box as its base, with printed pieces attached by screws or glue. The report says the design is not finished; the immediate aim is a rough first version, with later refinements possible.
AI as a Hands-On Design Assistant
The report offers a concrete example of AI supporting a maker workflow rather than producing a finished object on its own. For someone who finds it hard to turn an incomplete mental picture into a model, generated code can provide a starting point that is editable, testable and relatively quick to revise. In this case, the useful result was not a complete costume or prop, but a workable shape and a way to explore dimensions before committing to a final print.
The limitations are part of the story. Gemini ignored instructions about the placement of holes and flush side arms in one design, and added chamfers the maker did not want. Another request to split a fictional tool into printable subassemblies returned a breakdown the maker found unhelpful. The account suggests that AI can reduce some modeling friction, but users still need to check geometry, understand how parts fit and make decisions about fabrication.
That distinction matters to hobbyists deciding whether to use generative tools for physical projects. A code output can look plausible while failing a specific constraint; a print can reveal problems that a text prompt does not. The maker’s experience points to an iterative process—prompt, inspect, edit and test—rather than a single request that reliably yields a ready-to-build design.
beginner FDM 3D printer with build plate for cosplay props
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From Costume Plan to Printed Parts
The Halloween project is being developed under a limited schedule. The original idea of DIY power armor was set aside as too ambitious, and the delayed chest protector changed the costume plan. The maker instead intends to combine a Milwaukee hard hat, a flashlight attached with duct tape and a more distinctive handmade prop. The report does not describe a finished costume or a completed prop.
The maker has experience modeling familiar objects, but says new designs and multi-part assemblies are harder, especially when the overall concept is not yet settled. Gemini’s OpenSCAD output offered one route to testing shapes; the maker says another option would be to remodel components in Fusion or similar software for more direct control over dimensions and relationships between parts.
Some project details remain exploratory. Green air fittings in the design were printed from McMaster-Carr 3D files to check how quarter-inch-inner-diameter hoses might fit. The maker had not decided whether to use inexpensive metal fittings, and described the test as a way to avoid ordering an unnecessary range of parts. The prop’s fictional purpose was also undecided: the maker was considering a detector, while the son would ultimately name it.
“The AI made a lot of mistakes, and I’m also learning how to better constrain dimensions in my prompts.”
— The maker, in the ToolGuyd report
As an affiliate, we earn on qualifying purchases.
Design and Assembly Still in Progress
The report does not confirm whether the prop will be completed before Halloween, how the final version will look or which materials will make it into the build. The current design may change after low-density test prints, and the maker says a later version could be modeled in Fusion or similar software. It is also unclear whether the maker will use metal air fittings, how the outside wires will be attached, or what name the son will give the prop.
The source is a first-person account of one project, not a controlled test of Gemini or a comparison of AI design tools. It does not establish how often the system produces usable models, whether other users would get similar results, or how much time the process saves compared with manual modeling. The maker notes that Gemini Pro timed out during prompting and temporarily shifted to a lower AI tier, but gives no precise account of the interruption’s effect on the schedule.
OpenSCAD 3D modeling beginner guide
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
A Rough Version Before Halloween
The maker’s stated next step is to print a low-density version of the handle and continue testing the prop’s parts. If those tests work, the plan is to print components in a final color, likely black, and attach them to a small wooden box using screws or glue. The maker may manually remodel some pieces if that gives better control, though the report says a full redesign is unlikely for this build because of the time limit.
The immediate goal is a usable rough first version that the maker and his son can build together. Further refinement is a possibility, not a confirmed outcome. The final assembly, appearance and completion date remain dependent on the test prints and remaining work.
EVA foam sheets and cosplay craft supplies
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the maker building?
A Fallout-inspired Halloween prop to accompany a costume his son is developing. The exact fictional function and final name have not been decided.
How is Gemini being used?
The maker used Gemini to help create OpenSCAD code for shapes and parts, then adjusted parameters and tested the designs with 3D prints.
Did the AI produce a finished design?
No. The report describes incomplete models and errors, including missed instructions about holes and side arms. The maker still has to revise, print and assemble the parts.
Will the prop be ready for Halloween?
That is not confirmed. The maker’s goal is to produce a rough first version before Halloween, but the project was still in progress when the report was published.
What remains uncertain about the build?
The final design, materials, assembly method and completion date are all unsettled. The maker may revise components in other 3D software after testing, but has not confirmed that plan.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
