Search papers, labs, and topics across Lattice.
The paper introduces Creative Pre-trained Transformer (CPT), a decoder-only language model fine-tuned to predict visual style attributes in design templates represented in a novel Creative Markup Language (CML). CML captures canvas-level structure, layout, and element-level details, enabling CPT to generate editable design variations. Experiments demonstrate CPT's ability to generate contextual color and font variations, and adjust layouts while preserving design principles, offering fully editable design documents.
Forget static images – this system generates *editable* design variations by predicting style attributes, opening up new possibilities for personalized and scalable creative workflows.
Designing visually diverse and high-quality designs remains a manual, time-consuming process, limiting scalability and personalization in creative workflows. We present a system for generating editable design variations using a decoder-only language model, the Creative Pre-trained Transformer (CPT), trained to predict visual style attributes in design templates. At the core of our approach is a new representation called Creative Markup Language (CML), a compact, machine-learning-friendly format that captures canvas-level structure, page layout, and element-level details (text, images, and vector graphics), including both content and style. We fine-tune CPT on a large corpus of design templates authored by professional designers, enabling it to learn meaningful, context-aware predictions for attributes such as color schemes and font choices. The model produces semantically structured and stylistically coherent outputs, preserving internal consistency across elements. Unlike generative image models, our system yields fully editable design documents rather than pixel-only images, allowing users to iterate and personalize within a design editor. In experiments, our approach generates contextual color and font variations for existing templates and shows promise in adjusting layouts while maintaining design principles.
