NXTStream
off air
01

Contribution in

what the venue hands us

transportsrt://0.0.0.0:9711
picture1280×720p30
frame rate—
realtime—
dropped—

Listening. Start the feed, or push SRT to the port above from any encoder.

02

ASD spine

simulated

what we know is happening

Round 2 underway
Round 2 winneropens
round clock 00:00:00:00 · 1/7 events

In production this is the live event stream the trading desk already produces. Only the source of the rows changes.

03

Split

what leaves here

video renditions1 (passthrough)
audio renditions0
video reloads—
delivered nowmeasured
video—
audio × 0—
this pipeline—
six separate streamsderived—
video holdmodelled
1450 ms

Video is held so every language lands on one timeline. Miss this and the Japanese call arrives after the punch that caused it.

now hearing— · one video element · 0 loads since start

Routing matrix

1 video → 0 audio

    Video is passed through, never re-encoded. The bandwidth cost of 0 languages is measured in stage 03, not claimed here.

    What we knew, and when

    simulated
    • 00:00:00:00Round 2 underwayRound 2 winner — opensspoken

    Transmission log

    measured
    • Quiet.

    What is real here

    • The SRT contribution feed, from a separate encoder process
    • The ingest, the mux, and every number in stage 01
    • One video rendition and six audio renditions in one HLS master
    • Language switching with no video reload
    • The picture is synthesised, not a camera in a venue
    • The event spine is scripted, not observed
    • Commentary is pre-rendered TTS, not live ASR and translation
    • Latency budgets are modelled from the design, not measured