Abstract

Speech disfluency and low verbal confidence affect many autistic children and young people. Traditional speech therapy can be repetitive, stress-inducing, and poorly adapted to individual needs. Many existing tools rely on cloud-based processing, introducing privacy risks, connectivity constraints, and long-term reliability issues. There is a clear need for an accessible, engaging, and fully offline approach to speech development.

ReadingStar 2.0 is an AI-powered, offline, karaoke-style application for music-driven speech practice. It generates original lyrics and musical backing tracks using locally hosted generative models, with synchronised audio and text to guide users. A teacher-controlled supervisor mode allows adjustment of difficulty, speed, and feedback, while positive reinforcement encourages participation without penalisation. The system runs as a standalone Windows executable requiring no internet connectivity.

ReadingStar 2.0 delivers engaging and motivating speech practice through music-based interaction, increasing confidence and verbal participation. Building on positive classroom feedback, the project demonstrates the impact of offline AI and inclusive interface design in educational settings. The system was showcased at the BETT education technology conference, highlighting its relevance to real-world classroom deployment and ethical, resilient technology use.

Video Demo

This is our portfolio video walking through the usage and development of ReadingStar 2.0

Meet the Team

Project Timeline

The GAANT chart below shows the timeline of the project.

Our Partners