1. Customisable Virtual Audience
Avatar, environment, audience size, attire, and reaction style can be reconfigured before every session.
Public Speaking in VR is a customisable VR application for reducing foreign language anxiety — validated in a peer-reviewed Frontiers study, deployed in English classrooms at LASALLE Singapore, and built around repeated immersive practice rather than one-off exposure.
The project began as my final project in the MSc Virtual and Augmented Reality programme at Goldsmiths, University of London, developed in collaboration with LASALLE College of the Arts in Singapore. The goal was to design a VR-based English presentation training application to support students who experience anxiety when speaking in a foreign language.
Originally a semester-long project, it was extended for two years with support from the Goldsmiths–LASALLE Partnership Innovation Fund (PIF). Over this extended period, the system was further developed, tested, and academically refined — leading to multiple international conference presentations, peer-reviewed journal submissions, and a total of £10,000 in grant funding.
The application offers immersive, customisable environments where students can practice giving English presentations to a virtual audience, helping them build confidence and fluency in a low-stress setting. Beyond the lab, it was integrated into actual English language classes at LASALLE College of the Arts in Singapore — instructors used it as a supplementary tool during lesson time, giving students repeated in-class practice before speaking in front of peers.
Foreign language anxiety is one of the most persistent barriers in language education. Many students understand the material but freeze when asked to speak in front of others — especially in a second language, where fear of judgement amplifies every mistake.
Traditional classroom practice offers limited repetition and high social pressure. Students rarely get enough low-stakes exposure to build fluency, and instructors struggle to simulate realistic audience conditions at scale. The research question became: can repeated exposure to a customisable virtual audience reduce that anxiety without replacing human instruction?
Public Speaking in VR places learners at a virtual desk where they customise their avatar, audience, and environment before each two-minute presentation. Built in Unity — initially for Meta Quest 2 with hand tracking, and later on Meta Quest Pro for gaze tracking — the experience uses a first-person virtual mirror, a desk-mounted UI panel, and a responsive virtual audience whose size, attire, and reactions the learner controls.
Unlike many prior VR public-speaking studies built around identical repeated sessions, this system was designed so participants could adapt the difficulty each time — choosing a familiar classroom over a conference room, adjusting the audience size, or switching from neutral to positive feedback as confidence grew. That customisation mirrors systematic desensitisation: learners control their exposure, practice at their own pace, and gradually move towards more challenging settings. AI-driven audience members, hand tracking, and gaze tracking combine to create interactions that feel socially real — enough to trigger practice anxiety, but safe enough to encourage return visits.
The application separates the learner-facing VR experience from the research evaluation pipeline, so classroom practice and academic measurement can run in parallel.
Avatar, environment, audience size, attire, and reaction style can be reconfigured before every session.
Six repeated VR sessions produced significant reductions in self-reported FLA and measurable gains in speech fluency and clarity.
Hand tracking, gaze tracking, virtual mirrors, and avatar choice support embodied, learner-controlled practice.
Used during actual English language classes at LASALLE Singapore — not just in controlled lab studies, but as part of everyday lesson practice.
The system was evaluated through in-depth data collection and analysis, contributing to academic research and real-world outcomes.
Since childhood, public speaking was something I found emotionally difficult, especially in English as a second language. During the short time I lived in Canada as a child, the pressure of presenting in another language while many people were watching once made me cry during a presentation.
That memory stayed with me. At Goldsmiths, it became the motivation behind this project: I wanted to design, develop, and evaluate a VR public-speaking application that could test whether repeated immersive exposure might reduce foreign-language anxiety for people like me.
Building the system and running the study showed me that technology can support confidence and participation, but only when it is carefully designed around the learner’s own voice, choices, and agency. Customisation was not just a feature; it was how participants made the space theirs before speaking. That principle — designing for the person, not just the problem — is what I still carry into every project I build.