diff --git a/quest/m1/push-to-talk/README.md b/quest/m1/push-to-talk/README.md index 1ff239eec43..881b9ea9a9d 100644 --- a/quest/m1/push-to-talk/README.md +++ b/quest/m1/push-to-talk/README.md @@ -93,6 +93,10 @@ in [Container reads drop backfill](/quest/m1/read-backfill.md). The runner fails on a head that is received but never decoded, so the proof catches this. A complete set of runs finishes this epic. +## Required + +- [Native playback in the acceptance proof](/quest/m1/push-to-talk/native.md) - `just test push-to-talk` also runs a `moq play` listener, held to the same no-lost-head, comfort-noise, and no-underrun checks + ## Related - [Container reads drop backfill](/quest/m1/read-backfill.md) - changes how the ordered read treats a group below its cursor, where this head was lost diff --git a/quest/m1/push-to-talk/native.md b/quest/m1/push-to-talk/native.md new file mode 100644 index 00000000000..191d2fc370b --- /dev/null +++ b/quest/m1/push-to-talk/native.md @@ -0,0 +1,33 @@ +# [S] Push-to-talk's acceptance proof covers native playback + +## Goal + +`just test push-to-talk` (`test/audio-quality/ptt.sh`) runs a native +`moq play --metrics` listener beside the browser one, on its own session +through the same local relay, and holds it to the epic's correctness checks: +every utterance is heard from its first frame, each muted span's declared +silence plays as comfort noise, and no press causes an underrun or stall. + +Native playback learned to fill declared silence with comfort noise in +[moq-dev/moq#8010](https://github.com/moq-dev/moq/pull/8010), but only unit +tests cover it. The audio-quality harness's `pulse` rows grade native playback, +never through a mute and unmute. + +Non-goal: a native latency budget. +[Press-to-audible latency](/quest/m1/push-to-talk-latency.md) stays +browser-only, as its Goal records. + +## Plan + +Decided 2026-11-06 (maintainer, /quest-complete while queueing +[moq-dev/moq#8010](https://github.com/moq-dev/moq/pull/8010)): ✅ plan it +as a child of [Push-to-talk](/quest/m1/push-to-talk/README.md), checking +correctness only. Not chosen: folding it into the latency quest and budgeting +native latency; dropping it because unit tests and the `pulse` rows suffice. + +`moq play --metrics` already reports underruns and needs `--video off`. Run +it under the same ideal-path conditions as the browser listener (no shaping +or loss), and report its rows in the same `summary.json` so one press's +browser and native results sit together. If the native metrics cannot tell +a lost utterance head from comfort noise, add what it needs to `--metrics` +rather than inferring it in the script.