UK universities win £8.45m to build an AI sign language model with the Deaf community

A consortium of British universities won £8.45 million in January 2025 to build SignGPT, an artificial intelligence model designed to translate between spoken and written English and photo-realistic British Sign Language video. The five-year project is funded by the UK's Engineering and Physical Sciences Research Council and led by the University of Surrey, working with the University of Oxford and University College London's Deafness Cognition and Language Research Centre.

The technical challenge is steep. Sign languages are full natural languages with their own grammar and spatial structure, not simply signed versions of English, and translating fluidly between speech and lifelike signing video remains unsolved. The project plans to assemble what it describes as the world's largest sign language dataset, develop a sign language large language model, and release open-source toolkits, data-annotation tools and web-based demonstrations so others can build on the work.

“45 million in January 2025 to build SignGPT, an artificial intelligence model designed to translate between spoken and written English and photo-realistic British Sign Language video.”

Crucially, the Deaf community is meant to be a collaborator rather than a subject. The project includes representation from the Royal Association for Deaf People, reflecting a hard lesson from earlier accessibility efforts, where technologies such as signing gloves were designed without deep Deaf involvement and failed to reflect how sign languages actually work. Professor Richard Bowden framed the goal as ensuring the Deaf community is not left behind in the AI revolution.

A grant is a beginning, not a finished product, and several caveats apply. The hardest parts of the research lie ahead, generative video can produce errors, and no automated system should be assumed to replace qualified human interpreters for high-stakes situations such as medical or legal settings. But the combination of serious public funding, leading research groups and genuine community partnership makes this a credible, well-grounded effort to extend the benefits of language AI to people too often left out.