What we're shipping, what broke, what we learned. No marketing posts — just engineering notes for people running video infrastructure.
Honest operational lessons on which FFmpeg presets you can trust under load — psy-rd, lookahead, threading, GOP discipline, partition control. The defaults that bite, the settings worth pinning, and why preset stability matters across encoder versions.
A buyer's checklist for evaluating any video orchestration platform — the seven questions that surface what's actually going to bite you in production. Vendor-neutral, decision-tree style, written from the perspective of teams running real broadcast and OTT workloads.
An honest decision framework for engineering teams choosing between rolling their own FFmpeg-as-API, buying a managed transcoding API (MediaConvert, Bitmovin, Mux), or renting orchestration-as-a-platform. The economics, the lock-in math, and which fits which workload.
An honest decision framework for broadcasters and OTT engineering teams choosing between building video workflow orchestration in-house and buying it. The build case, the buy case, and why most teams end up doing both.
How to run FFmpeg in Kubernetes at production scale — the four patterns we've watched teams climb, where each breaks, and why a video transcoder Kubernetes operator earns its keep above ~50K jobs/day.
Honest comparison of Bitmovin Encoding and Mux Video for engineering teams choosing between them. Where each wins, where the decision flips, and the gap that neither one closes today.
When self-hosted video transcoding actually beats AWS MediaConvert. Cost math, hardware, ops trade-offs, hybrid architectures — without the vendor slide-deck framing.
A thesis on modeling video processing as a graph instead of a sequence. The trade-offs, the failure modes scripts hit at scale, and what graph structure buys you in production.
What FFmpeg-in-production actually demands — the queue patterns, retry semantics, and audit-trail design that get a single binary to behave like infrastructure.
What we're shipping, what broke, what we learned. No marketing posts — just engineering notes for people running video infrastructure.