Foundations for graphic design AI research
At Lica, we don't believe creative AI is about re-prompting. Our goals for building scalable design models are simple:
Research at Lica defines AI for graphic design, pushing the boundaries of how machines understand and create visual content. Central to this is the layout—the organized arrangement of text, images, and graphics that guides attention, communicates ideas, and creates visual harmony.
We build intelligent systems that generate layouts, suggest improvements, and collaborate with humans in real time, blending computer vision, generative AI, and human-centered design to make creativity faster, smarter, and more accessible.
Our ultimate vision is the graphic design co-pilot: an AI that truly understands design and works alongside creators to bring ideas to life. It anticipates needs, suggests improvements, and balances guidance with control, so designers can perfect every detail.
Seek perfection and hate re-prompting? We do too.
We build models that understand how elements are positioned, structured, and aligned, enabling designs that feel balanced, readable, and intentional. This spatial intelligence forms the foundation of visually compelling layouts.
Our models plan, generate, and refine layouts, while adapting to different styles through learning. This allows designers to explore new compositions or quickly edit existing ones with precision.
We train AI to reason about vector graphics, enabling text- and image-to-SVG generation that is scalable, editable, and brand-aware. This unlocks design workflows that go beyond pixels to structured, semantic graphics.
Our systems learn and adapt to user preferences, assets, and brand identity—colors, fonts, visuals, and beyond. This ensures every design output feels authentic, aligned, and true to its creator.
We extend static designs into motion by generating animations for text, images, and SVGs. This brings layouts to life with fluid, dynamic storytelling.
Work at lica.world? Sign in to claim this org and manage its profile.