Editable Visual Design
The paper proposes an agent workflow that outputs editable HTML/CSS designs instead of flattened images.
The system uses a VLM for planning and aesthetic judgment, then calls an image model to generate standalone visual assets. It writes native HTML/CSS, checks rendered feedback, and iterates on the design. The authors say the result keeps layers separate and text real, so users can drag and adjust elements in a GUI. They report validations on posters, infographics, and related design scenarios. HF Daily Papers' note
The system uses a VLM for planning and aesthetic judgment, then calls an image model to generate standalone visual assets. It writes native HTML/CSS, checks rendered feedback, and iterates on the design. The authors say the result keeps layers separate and text real, so users can drag and adjust elements in a GUI. They report validations on posters, infographics, and related design scenarios. HF Daily Papers' note
score 5