Large-session video infrastructure for online course platforms

An online course provider schedules a live cohort session for 120 students and discovers, mid-session, that their conferencing layer starts dropping participants and freezing cameras around the 50-person mark. They've just paid for a platform that explicitly promises live learning — and it falls apart at the moment that matters most. The problem is that the major conferencing vendors built their video mixing and relay infrastructure for general-purpose business meetings, not sustained high-participant educational sessions. Fixing this properly would require architectural changes to how they handle simultaneous video streams — work that cannibalizes their existing customer base who mostly don't notice the failure because they're in 10-person standups.

Current conferencing layers handle participant overflow by degrading gracefully — pixelating cameras, dropping to audio-only, or silently kicking users. That's fine for a sales call. For a training session with 150 attendees where a trainer is being evaluated on delivery, or a virtual university course where attendance is graded, a silent drop is a serious operational failure. Users describe software 'freezing during a virtual training session with 150 attendees' and 'crashing for several attendees whenever more than 6 had video on' — these aren't power-user edge cases, they're the core use case for anyone running live online education.

The real cost is concrete: a corporate L&D team running a mandatory compliance training has to reschedule, which means rebooking a facilitator, notifying 150 employees, and eating the soft cost of disrupted schedules. An online school loses credibility with enrolled students. Neither buyer has a good workaround — downgrading to audio-only defeats the purpose of live instruction, and splitting into smaller breakout rooms requires a facilitator per room they don't have.

This is a business and not a feature because the conferencing vendors won't fix it — not because they can't, but because the marginal revenue from enterprise training customers doesn't justify the infrastructure investment when most of their revenue comes from smaller-meeting SaaS subscriptions. A specialized layer that handles video relay, stream prioritization, and graceful degradation specifically tuned for 50–500 participant educational sessions, licensed to LMS and course platform vendors as a white-label embed, creates a recurring infrastructure contract with companies whose entire product depends on it working.

What to build

Build a WebRTC-based video relay service, deployable as an embeddable SDK, that manages per-participant bandwidth allocation, server-side stream compositing for large rooms, and graceful camera fallback logic — exposed via a single API that LMS platforms drop in to replace their current conferencing embed.

Where to start

Target open-source LMS communities first — platforms running OpenMeetings or self-hosted conferencing already have no vendor lock-in and a documented performance problem at scale, making them willing to evaluate a drop-in replacement with no contract risk.

The hard part

You need large-scale load test data to prove reliability before any LMS will swap out their existing video layer, but you can't generate that data without real traffic — meaning your first customer is taking a risk you can't fully quantify yet.

How it makes money

Usage-based pricing per participant-minute above a free tier threshold, with a flat monthly infrastructure fee for platforms that need guaranteed capacity; contract size scales naturally as the LMS grows its live session usage.