AI image models are great at generating visuals, but they still struggle with exact text, typography, and layout.
I built VisualSpec, an open-source tool that separates these jobs:
- AI generates the image/background
- SVG/code renders the exact text and layout
- Visual QA checks margins, dimensions, and layout issues
- MCP lets Cursor use VisualSpec directly
Demo 
Run it:
npx @utkarshx27/visualspec mcp
GitHub:
npm:
I’m looking for feedback from Cursor/MCP developers, and I’d also love contributors interested in TypeScript, AI image generation, templates, and visual QA.
Would love to know what you think or what you would want VisualSpec to support next.