A developer says Anthropic's Claude Opus 5.5 can turn a single reference picture into a complete three.js "teardown" model, with every shell, gear and hose written as code, and has packaged the full workflow into one reusable Claude skill.
Claude Opus 5.5 · Three.js teardown demo
One reference picture, a full 3D teardown written as editable code
In a developer's demo, Claude Opus 5.5 turned a single image into a three.js scene in which every shell, gear and hose is its own block of code. The workflow was then packaged as a reusable Claude skill.
1
reference image sets the overall look
3
part types coded separately: shells, gears, hoses
1
reusable skill packages the whole workflow
How the workflow runs
Steps as described in the demo
STEP 1
Read one reference image
→
STEP 2
Set the overall look of the object
→
STEP 3
Write three.js code for each part
→
STEP 4
Render and compare with the picture
↺ The full sequence is packaged as one Claude skill, so it can be run again without rebuilding the prompts.
Why the approach stands out
Each shell, gear and hose stays a separate, inspectable element
Parts can be modified, repositioned or reused
Output runs in a standard web environment via three.js
What the demo does not show
No independent benchmarks or model comparisons
No measured accuracy of the match to the reference
No check on how hidden internal parts were inferred
Open questions
Will the results hold for more complex machinery, and will reusable skills become a common pattern?
Product visualization
Engineering education
Technical illustration
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...