Start with distance and direction
Place the microphone consistently and speak across rather than directly into sensitive capsules when plosives are a problem.
Adjust input gain gradually
Raise system or interface gain until normal speech is clearly above noise while loud syllables remain below clipping.
Do not fix clipping in software
Once the input clips, lowering playback volume cannot recover the flattened waveform. Reduce gain at or before capture.
Listen to a real sample
Meters cannot reveal every echo, buzz, keyboard sound, or processing artifact. Playback remains the practical final check.
Read RMS and peak as different evidence
RMS approximates sustained signal energy, while peak shows the largest instantaneous sample in the current analysis window. Normal speech can have a moderate RMS value and much higher peaks. A near-zero RMS suggests silence, a disconnected source, or input that is too quiet; repeated peaks near the digital maximum suggest clipping risk. These browser measurements are relative full-scale values, not calibrated sound-pressure levels. Use them to compare settings on the same device, then confirm intelligibility through the local recording.
Retest one variable at a time
Keep speaking distance and voice level consistent, change either hardware gain, system input level, or microphone position, and record another short sample. Browser-requested echo cancellation, noise suppression, and automatic gain control may be ignored or implemented differently by the operating system, so the active-track settings are reported when available. Stop capture before switching applications, and compare in the conferencing application too: it may add its own processing after the browser test. Do not infer room safety or hearing exposure from this uncalibrated meter.