Dizzy is a product architecture concept for voice-driven UI design. Instead of generating images, it streams voice input to an agent that builds a semantic command buffer. This buffer translates intent into structured operations for the Figma MCP, allowing the system to progressively modify native Figma designs. Later commands can intelligently update earlier decisions rather than creating entirely new assets.
Manipulates native Figma objects via Figma MCP instead of generating images.
Maintains structured design state to allow progressive and contextual modifications.
JEV/Streaming Voice → Agent Intent → Design Operations → Figma MCP.
Expanding into an agentic product design loop integrating visual comparison and code generation.