MAX_WIENS
PROJECT_LOG
↩ RETURN TO HUB⚡ FEATURED_LOG // Soft Skills VR Training // Varjo Headset
This was a genuinely unique project developed for the University of Manitoba, this experience was built for the Varjo headset, which pushed the visual fidelity and eye-tracking capabilities of the simulation to a level consumer hardware simply can't match.
Users step into the role of a university professor navigating tense office conversations with a virtual student. The core of the experience is a branching narrative dialogue system I built that gives users real choices between empathetic and dismissive approaches, and then lets them feel the consequences of those decisions play out in real time.
What made this build technically interesting was the advanced eye-tracking integration using the Varjo's native hardware. I wired up attention and stress-response measurement running live during each interaction, surfacing real-time data on the user's eye contact percentage alongside their per-action response times — giving instructors and researchers a genuine performance analytics layer on top of the training experience.
The characters themselves are powered by AI-driven response systems with synchronized audio and naturalistic animations that react dynamically to whichever dialogue branch the user selects. The pilot phase demonstrated measurable improvements in both recall of stressful scenarios and decision-making quality among participants.
Core System Architecture:
- Branching Dialogue Engine: Interactive narrative system that branches on user choice between empathetic and dismissive response paths, with each selection driving unique AI character reactions and conversation outcomes.
- Varjo Eye-Tracking Integration: Native Varjo hardware eye-tracking pipeline measuring live attention and stress responses — outputting real-time eye contact percentage scores per interaction session.
- Response Time Analytics: Per-action decision timing system that logs and displays user response latency across all dialogue actions, giving instructors a granular performance data layer for post-session review.
- AI-Driven NPC System: Character models with dynamic AI response logic, synchronized audio delivery, and naturalistic animation playback that adapts to the user's specific dialogue selections.
- Performance Feedback Panel: Post-session debrief screen surfacing eye contact percentage, action-by-action response times, and qualitative feedback on the effectiveness of the user's chosen dialogue approach.
