Modular Sign Language synthesis using high fidelity 3D avatars

Sign language is the primary means of communication for the Deaf and Hard of Hearing (DHH) community. As the production of digital media grows, relying solely on human interpreters is not always feasible, creating the need for automated, scalable solutions. Signing avatars offer a promising approach, but present significant challenges due to the linguistic and expressive complexity of sign languages. This thesis presents a prototype system that dynamically generates Italian Sign Language (LIS) sentences on a MetaHuman 3D avatar by smoothly blending individual signs recorded through motion capture. The system was developed at the Rai Centre for Research, Technological Innovation and Experimentation (CRITS), extending their existing platform for generating fixed sentences. My contribution enables users to construct custom sentences from a dictionary of signs that can be easily expanded, producing output evaluated as natural and intelligible by certified signers and DHH users. The work demonstrates the feasibility of modular sign synthesis using high-fidelity avatars, and outlines future directions including automatic text-to-gloss translation and the integration of non-manual features, paving the way toward an accessible, on-demand sign language translation system.