Blog

Behind the Scenes: Engineering Playmist

By the Playmist Engineering Team — reading time: 6 minutes

When we set out to rebuild our video delivery pipeline last year, we had three goals: cut start-up latency in half, support live events without degrading on-demand quality, and do it all without adding a single line of duplicated code to our apps. This is the story of how we got there.

The old pipeline

Our previous stack was a patchwork of origin servers, a single CDN, and a custom transcoder that had served us well for three years but was starting to show its age. The average time-to-first-frame on a cold client was 3.4 seconds — acceptable, but not the experience we wanted.

Rethinking the manifest

The biggest win came from switching to a multi-CDN, edge-signed manifest. By moving token generation to the edge, we eliminated a round-trip to our origin and cut initial buffer time to under one second for 85% of sessions.

"Every millisecond we shave off the start-up feels like magic to the user, and it's earned by real engineering work down the stack."

What's next

We're now rolling out per-title encoding with content-aware bitrate ladders, which should further reduce bandwidth without visible quality loss. Expect a deeper write-up on that later this quarter.

Interested in the kinds of problems we're solving? We're hiring.