Recorder diagnostic
Separates microphone input, recorder behaviour, file contents and playback. Everything stays in this page's memory. No audio leaves the device.
1 · This run
Everything else is held constant: the same microphone request,
no bitrate set, { mimeType } passed unconditionally, the same stop
path and the same analysis. Diagnostic variants are not the
application's settings.
Fill these in before pressing Start. They are recorded when a recording begins; later edits apply only to the next run. Use the button below only to fix a typo in the run you just finished.
2 · Record — 10 to 15 seconds is enough
Off by default, deliberately. The meter taps the capture stream through an AudioContext, which on a tablet can itself alter the audio session. The original symptom must be observed before anything changes that session, so leave this OFF for the first run and turn it on only once you have a baseline. Whether it was on is recorded with every run.
Speak normally and watch the level move. If the meter stays flat while you talk, the microphone is not delivering signal and nothing downstream can.
3 · Play it and listen — do this FIRST
Record this before decoding. No measurement on this page can substitute for it: a decoder can produce a signal that is noise, corruption or incomplete audio.
| Status | no recording yet |
4 · Decode and analyse no run yet
The analysis context is closed when the decode finishes and the outcome is recorded, and the next run re-reads that context's state before recording rather than assuming it is gone. That is a record of two specific actions — it is not a claim that every audio-session effect has reset.
Do this after listening. Decoding happens in this page and answers one narrow question: whether the decoder produces non-silent samples from this file. It cannot tell you the audio is correct or intelligible, and a decode failure is reported as a failure — never as silence, never as success, and never as proof that the device cannot decode the format.
5 · Measurement
| Status | no recording yet |
6 · Report 0 runs
Diagnostic text only — never audio. Each run is kept separate. Reloading clears everything.
Provenance and limits
The two ORIGINAL profiles are transcribed from application source at revision
b5fe49566b647cd5e9d22fd164a615eefd1a534b (2026-09-21)
They match that source revision. This page does not independently verify production behaviour. The DIAGNOSTIC variants are deliberately not the application's settings.
No profile here reproduces the Live Conversation recorder's audio source: that recorder wraps a live room's already-published call track, which no standalone page can substitute for. Sizes measured here are indicative of settings, not equivalent to the live-room track, and do not replace a real classroom test.