01

Recreated from trailer footage alone in under 48 hours

Steam Controller — built in a day, mistaken for the real product

Year
2025
Type
Personal Project · Hard-Surface Modeling
Role
3D Artist

When Valve announced the Steam Controller, no physical unit existed yet — only trailer footage and promo images. I set myself a challenge: recreate it entirely from reference, at photorealistic quality, in under two days. The result was mistaken for the actual product by thousands of people.

No CAD, no measurements, no physical reference

The controller had been announced but not shipped. Only a 2D trailer and a handful of promotional screenshots existed. I had to infer proportions, depth, button placement, material finishes, and ergonomic curves purely from visual reference — no CAD data, no blueprints, no physical unit to measure.

Every angle had to be reconstructed from limited views. A single misinterpreted perspective could throw the entire model off.

Reference materials for the Steam Controller

Pure Blender hard-surface workflow

No shortcuts, no photogrammetry — just reference images pinned to the screen and a Blender hard-surface workflow from scratch.

Day 1 was pure modeling: blocking out the main body, then layering in every detail — thumbsticks, D-pad, face buttons, triggers, grip contours, the touchpad surfaces, the Steam button, the LED indicator. Every component was modeled separately for accuracy.

Day 2 was shading, lighting, and rendering. I set up a studio-style HDRI environment, assigned PBR materials (soft-touch plastic for the grips, glossy finish for the face, matte for the buttons), and dialled in the lighting until it matched the tone of the official promo material.

60,000+ people thought it was the real thing

Posted the final render to Reddit. The response was immediate — and unexpected. Thousands of comments from people asking where to buy it, how much it costs, and when it ships. I had to clarify in every thread that it was a 3D model, not a photograph.

The post accumulated 60,000+ views across multiple subreddits. No ads, no promotion — just the model speaking for itself.

This project validated something important: at this level of fidelity, the line between rendering and photography disappears. If you can make something that people mistake for a manufactured product, you can make a configurator that sells it.

Key Takeaways

01
Speed doesn't mean rushed
One day for modeling, one day for rendering — but the model held up at photorealistic quality. The key was having a clear mental plan before touching Blender. Every button, curve, and material was decided before the first vertex was placed.
02
Reference limitations force better observation
With only trailer footage and promo images, I had to study every frame carefully. That deep observation meant I noticed details most people miss — subtle bevel radia, material transitions, the exact falloff of the grip texture. Limited reference made the model better, not worse.
03
Photorealism is its own marketing
60,000 organic views came from one Reddit post. No ads, no promotion. The renders did all the work. If the model is indistinguishable from a photo, people share it because it feels like magic. That's the standard client configurators should aim for.

Process & Breakdown

Wireframe breakdown of the Steam Controller model1 / 2Clean topology — every edge has a purpose
48 hrs
From reference to final render
60K+
Reddit views — organic, no promotion
1:1
Proportional accuracy from limited reference

Final Render

Final photorealistic render

Back to portfolio

View all projects
Razershake — Technical 3D Artist