Summary
Creating 3D models in CAD software can be hard for many people, especially when they want to change parts later. This paper looks at how letting the computer understand design intent, parameters, and relationships helps generate CAD programs from text descriptions that are easy to edit afterward. The authors made a large dataset of text paired with CAD code to train and test their method. Their two-step approach first groups basic shapes into meaningful features, then adds adjustable parameters to make editing simpler without changing the underlying model. They found that focusing on semantic and parametric details improves the accuracy and editability of generated CAD programs.
What this means in practice
- •For product design teams: Generate editable CAD models from natural language descriptions for faster product concept iteration and customization.
- •For software developers: Build CAD tools that produce structured and reusable code, enabling downstream parametric editing without manual recoding.
Abstract
Text-to-CAD generation aims to make CAD modeling more accessible by generating executable CAD programs from natural-language descriptions. However, practical CAD workflows require more than executable geometry: generated programs should also expose reusable structure that supports later modification. We introduce intent-oriented text-to-CAD generation, where design intent is expressed as recoverable code-level signals: semantic features, reusable parameters, and lightweight parametric relations. Rather than changing the model architecture or decoding procedure, we investigate how supervision representation granularity affects both CAD generation and zero-shot intent-preserving editing. To enable this study, we construct IntentCAD-100K, a 100K paired text--code dataset built by converting low-level CAD construction traces into executable CadQuery programs. We further construct IntentCAD-Edit-1K to evaluate zero-shot intent-preserving editing from the same representation perspective. Intent2CAD uses a two-stage conservative lifting pipeline: semantic lifting rewrites reliable primitive groups into feature-level operations, while parametric lifting exposes lightweight parametric relations. Under controlled fine-tuning, semantic lifting mainly improves generation reliability and feature recovery, reducing IR from 2.39% to 0.22% and achieving 99.96% API Recovery, while parametric lifting makes reusable relation structure explicit and further improves relation-aware zero-shot editing. Datasets and code are available at https://github.com/JinboYang94/Intent2CAD.