Typus integrating Real Material Manufactures

Short description of the innovation study

The FFplus innovation study led by typus.ai develops a domain-specific generative AI framework for professional architectural visualisation. The project addresses key limitations of current generative AI tools, including inaccurate in terms of material representation and uncertainties around data provenance and copyright. The project combines licensed manufacturer material data, architectural 3D/BIM information and synthetic training data with customised generative AI models. This enables the generation of architectural visualisations that are more closely aligned with the underlying geometry and real construction materials. The study will develop and evaluate a functional AI prototype, establish a documented and copyright-safe data and training pipeline, benchmark the results against existing approaches, and demonstrate an initial integration into the typus.ai SaaS platform. The project is supported by access to European high-performance computing resources through the EuroHPC ecosystem.

Organisations involved: 

Main SME: TYPUS.ai

Technology expert: RWTH Aachen University - Chair of Computer Graphics, Geometry and Multimedia