WebSight: Large dataset for converting screenshots to HTML with Tailwind CSS and Sightseer model

AI Impact Summary

WebSight provides a large 2M+ example dataset of screenshot-to-HTML pairs to train vision-language models for converting visual UI designs into functional code. The v0.2 upgrade brings real images and Tailwind CSS, aligning generated HTML with modern frontend stacks and improving realistic code generation. The Sightseer model demonstrates end-to-end capabilities, including embedding image content into the resulting HTML, signaling a practical move toward rapid, low-code UI prototyping and faster handoff from design to deployable output.