Figma Launches VectorAI: On-Canvas Generative Vector Tooling

Design · 4 min read

VectorAI integrates a large multimodal model directly into Figma's editor to produce native vector outputs: scalable shapes, compound paths, gradients, and layer groups that behave like hand-crafted assets. Designers can prompt for specific styles, constraints, or accessibility considerations and tweak resulting nodes with standard vector tools.

Unlike previous raster-first generation tools, VectorAI's export preserves layer semantics so design systems, tokens and constraints remain intact. Generated assets come with suggested componentization and naming conventions to accelerate system adoption. Figma also includes a 'trace' mode to convert raster inspirations into editable vectors using the underlying model.

Figma says VectorAI is optimized for iteration speed, with a low-latency inference engine for quick refinements. The company will offer usage quotas for free accounts and expanded inference bundles for enterprise customers, alongside new governance settings for IP and output provenance.