Live video streaming, one-to-many

The problem

Customers needed one-to-many live broadcast for events, courses, and creator broadcasts. HLS alone had too much latency for interactive use; WebRTC alone scaled poorly to large audiences; viewer-side captions, chat sidecar, recorded VOD, DRM-aware paywall hooks, and CDN scaling all had to be wired together without dropping frames. A single failure mode killed the entire event.

The approach

We built a streaming platform with WebRTC for interactive sub-second latency tiers and HLS for adaptive-bitrate broadcast at scale, with a smooth handoff between the two. Viewer-side captions ride along with the stream; chat is a sidecar over WebSocket; VOD records at the origin; DRM hooks gate paywalled content; CDN scales the egress. FFmpeg handles transcoding at the origin without touching the live path.

Stack and engineering choices

• WebRTC + HLS hybrid
• Adaptive bitrate
• Viewer-side captions
• Chat sidecar over WebSocket
• Recorded VOD at origin
• DRM-aware paywall
• CDN egress scaling

Outcome

Events, courses, and creator broadcasts ship without dropped frames or audience overflow. Recorded VOD lives next to the live stream; paywalls work without leaking content; latency tiers match the use case (interactive vs broadcast). The platform absorbs audience spikes without reconfiguration.