Why one hundred recording chunks do not prove one hundred seconds
MediaRecorder chunk delivery intervals can be delayed. Measure elapsed recording time separately rather than multiplying a requested interval by the chunk count.
The journal
The viewing journal — page 54
Practical reading about streaming, devices and the decisions behind a better prepared setup.
A hidden browser caption track can still load cues
TextTrack hidden mode keeps cues active without drawing captions. Disabled mode can defer fetching the cue resource.
A dropped frame total needs a before and after comparison
Playback quality counters become more useful when compared over the same observation interval. A lifetime total does not identify the most recent visible hitch.
A video frame callback does not automatically measure camera to screen delay
requestVideoFrameCallback exposes media presentation timing. Ordinary video frame timestamps alone do not supply a verified original capture time.
Why faster browser playback can keep a voice at its normal pitch
The preservesPitch property controls compensation for playback rate changes. A shorter listening time does not require a higher pitched voice.
A browser speaker selector can fail even when system audio works
Selecting an audio output in a web player depends on browser support, permission and the specific device identity. System audio success does not prove the web selection request succeeded.