Image2LEGO: turn an image into a 3D brick model

Upload a reference image, explore an interactive model, follow its building steps, inspect the parts list, and export LDraw, CSV, or JSON files.

Image to brick model guide

Turn a picture into an editable 3D brick model

Start with a portrait, pet, building, vehicle, landscape, or an original sketch. Almost any subject can inspire a brick-design proposal; the model still needs your review before it becomes a physical build.

Start with a clear JPG, PNG, or WebP reference image. Choose the main view and describe the features that matter most, such as the subject's silhouette, colors, roof, face, or target size. Image2LEGO uses that reference to propose a three-dimensional brick arrangement that you can rotate and inspect in the builder above.

1. Review the model

Rotate the preview and check whether the large shapes match the reference. Image-based generation may infer unseen sides or simplify details; use a focused follow-up instruction when a distinctive feature is missing.

2. Inspect steps and parts

Open the assembly timeline and parts view to check the proposed build order, piece counts, colors, and structure. A convincing render alone does not establish that every part and connection is physically valid.

3. Export for further editing

Export LDraw to continue in compatible brick-design software, CSV for a bill of materials, or JSON for structured project data. Review the resulting inventory before ordering physical pieces.

What should you verify before building?

Check that each shape corresponds to a real part in the desired color, connections align, pieces do not collide, the structure has support, and the steps are possible to assemble. Image2LEGO output is an editable design proposal rather than a guarantee of part availability or real-world stability. Our seven-test buildability checklist explains each check with practical examples.

What published brick-generation research actually reports

In the 2025 BrickGPT paper, Ava Pun and colleagues report a StableText2Brick dataset with more than 47,000 brick structures derived from more than 28,000 unique 3D objects. The 2026 BrickNet paper by Peter Kulits and Cordelia Schmid reports 320,808 LDraw samples, 9,743 part variants, and 40,549,969 placed brick instances.

“we employ an efficient validity check and physics-aware rollback”
Ava Pun et al., BrickGPT (2025)
“it is the spatial relationships of the parts which define the whole”
Peter Kulits and Cordelia Schmid, BrickNet (2026)

Sources checked October 1, 2026. These figures describe the papers' research datasets—not Image2LEGO accuracy, performance, or implementation. Neither paper evaluates this website. Read our research explanation for the distinction.