Figma 2026.2 adds native multimodal AI assistant and vector-aware generation

Design · 5 min read

Figma 2026.2 rolls out a built-in multimodal assistant that combines image understanding, sketch interpretation, and text prompts directly in the editor. Designers can now drop photos or rough sketches into frames and ask the assistant to generate scalable vector icons, layout suggestions, and component variants.

Crucially for production teams, Figma introduced vector-aware generation that outputs editable SVG paths rather than flattened images. The company also added provenance metadata so generated assets carry source prompts and model versions, improving audit trails for design systems and legal compliance.

The update includes admin controls to limit use of external generative models and new guardrails for output filters. Early previews indicate designers appreciate the speed of vector outputs, though some request finer control over path simplification and boolean operations.