PlaybackTiming

PlaybackTiming = object

What the provider measured about the playback. Absent when it measured nothing. ‘firstAudioMs’ and ‘firstChunkMs’ count from the same origin: the moment the provider’s run() or stream() started handling the request, so their difference is the provider’s output latency. The span includes everything the provider does after it starts handling the request (queue wait, voice lookup, a cold model load) and excludes routing.

Indexable

[property: string]: unknown

Properties

durationMs?

optional durationMs?: number

How long audible playback lasted, in milliseconds. For a cancelled run, how long it played before it was stopped, when the provider reports it.

---

firstAudioMs?

optional firstAudioMs?: number

Milliseconds from the provider receiving the request until the platform’s first audible-output callback (for a system voice, its first word callback) or the first audio frame leaving the output. Absent when nothing reports it.

---

firstChunkMs?

optional firstChunkMs?: number

Milliseconds from the provider receiving the request until its first playable PCM chunk existed. Absent when the provider does not synthesize PCM itself.