XtendLive

Background

Most virtual event platforms in 2020 were glorified webinars. A speaker talked, attendees watched, and everyone left wondering why they didn't just send a recording. XtendLive wanted something different: 3D venues where people could actually walk around, visit exhibition booths, bump into each other in a hallway, and attend sessions, all inside a browser.

The idea was to recreate what makes physical conferences work. Not just the talks, but the wandering, the booth visits, the unplanned conversations. The platform needed to handle thousands of people in the same space at once, stream live video, and run on a normal laptop without installing anything.

Technical approach

React handled the frontend, paired with WebGL for the 3D rendering. We layered standard UI components (chat panels, session lists, attendee info) on top of the 3D environment so the interface felt familiar even though you were navigating a virtual space.

Node.js ran the backend and managed WebSocket connections for everything happening in real time: people moving through the venue, chat messages, poll responses, session state changes. With thousands of concurrent users, the event-driven model was the only sensible choice.

Video streaming ran on WebRTC. Keynotes used an SFU architecture to fan out to large audiences, while one-on-one meetings in networking areas used direct peer connections. Adaptive bitrate kept playback smooth across varying connection speeds.

On the infrastructure side, AWS handled everything. EC2 for compute, CloudFront for distributing 3D assets and video, S3 for storage. PostgreSQL was the primary database for user sessions, venue configs, exhibition data, and analytics.

Implementation details

The venue builder let organizers lay out their conference in 3D: place exhibition booths, set up session rooms, brand the common areas, add networking lounges. Each element linked to actual content. Booths had company info and downloadable materials. Session rooms connected to live or pre-recorded streams. Networking areas opened spontaneous video calls between attendees.

Navigation used simple point-and-click controls with a minimap for larger venues. Getting this to run well on mid-range hardware took real work. We used level-of-detail rendering so distant objects loaded at lower fidelity, lazy-loaded assets as people moved through the space, and compressed textures aggressively. Everything ran in the browser with no plugins.

The engagement engine processed thousands of interactions at once across the 3D space. People visited booths, grabbed virtual brochures, answered live polls during sessions, and exchanged contact details through virtual business cards. Chat worked at two levels: venue-wide announcements and proximity-based conversations that only reached people in the same area.

We ran this project for four years, so the DevOps setup had to be solid. Infrastructure as code, automated scaling for traffic spikes during big events, and monitoring that covered both the application layer and 3D rendering performance.

Outcomes

The platform ran virtual conferences and exhibitions for organizations worldwide from 2020 through early 2025. The 3D format changed how people behaved at events. Attendees actually explored exhibition areas and spent time networking, much closer to physical event patterns than what you see on flat webinar platforms. Exhibitors got something they couldn't get elsewhere: a spatial way to show products and talk to qualified leads, rather than hoping someone clicks a link in a chat sidebar.