Best workflow for AI-generated UI mockup to real DOM web page?

Source: Dev.to

I am prototyping a mobile web page from a high-fidelity AI-generated visual mockup. The visual style is complex: wooden tabletop, paper cards, fabric, soft shadows, small illustrations, and textured buttons. The final page cannot be a flat image with transparent hotspots. It needs real DOM text and real controls because the content is dynamic: participant names can change; card labels and descriptions can change; buttons need click/focus states; the page needs to work on mobile web. What I tried: CSS-only recreation: interaction works, but visual quality drops a lot. Full-page image background plus DOM overlay: visually closer, but DOM text does not align reliably and the asset is not maintainable. Text removal / inpainting from the mockup: creates artifacts, especially around paper/card areas. What I am looking for: A mature workflow for converting a complex bitmap mockup into layered web assets. Which parts should be done in Figma, Photoshop, or another tool. Whether AI layer extraction tools are good enough for production. How to keep all text/buttons as real DOM. How to verify visual similarity with screenshots or pixel diff. For a generic example, imagine a “Tea Tasting Journal” mobile page with: a warm desk background; a paper panel with participant cards; a grid of paper option cards; real buttons for “Start” and “Open Journal”; dynamic text over textured visual assets. Is the best practice still manual layer reconstruction in Figma/Photoshop, then DOM overlay, then Playwright visual regression? Or is there a more mature image-to-layer / design-to-code workflow? I am especially interested in workflows people have actually shipped, not only marketing demos.