Every low-latency streaming vendor at NAB or IBC will hit you with “sub-second latency” in demos. The reality we experienced running these solutions in production for interactive TV shows was quite different. This talk walks through the full evaluation process we went through, covering everything the vendors do not tell you upfront.
On the ingest side, tuning RTMP parameters – bitrate, codec settings, audio-video offset, portrait vs landscape – all impact latency. With RTMP ingest, we could never reliably get below two to three seconds with stable buffering. The SDKs for iOS and Android were enormous (60-100 megabytes for a video player), compatibility was inconsistent across OS versions, and Mobile Safari support was practically nonexistent at the time. Developer experience was rough: zip files instead of proper package distribution, sparse documentation, and no changelogs.
A critical insight for our use case: latency consistency mattered more than absolute latency. If everyone sees the same content at the same time – even with a two-second delay – that is fine for gambling or interactive formats. Only about 30% of the players we evaluated supported synchronization across devices. Advertising integration was completely missing from ultra-low latency players, forcing us to switch between a traditional player for ads and the low-latency player for content. Stream protection was basic at best, with no DRM support and limited authentication.
In production, the real problems surfaced: network ports blocked by corporate firewalls, CPU-bound decoding failures on older devices, battery drain (a 10-minute show drained 90% of an idle phone), poor foreground/background recovery, and no session IDs for debugging user complaints. Our advice: always show visual indicators when something fails – spinners, network warnings, anything – because a black screen just looks broken. Demand detailed metrics from your provider, not just green/red status, and insist on direct engineer access for live-show escalation.
Watch on YouTube – available on the jedi4ever channel
This summary was generated using AI based on the auto-generated transcript.