Every measurement behind your score, explained.

Key, fill, and background brightness measured directly across your frames, with a key:fill ratio — not just "looks a bit dark."

Checks whether your setup approximates a proper key/fill/background balance, the standard used in professional video lighting.

Flags genuine clutter or distraction, and suggests a blur or virtual background when it would help. Plants and tasteful decor are never counted against you.

Face position checked against rule-of-thirds framing geometry, headroom, and centering — plus posture and how tight or loose your framing is.

Resolution and encoding profile actually delivered, plus how evenly frames arrived — stutter and freezing show up here.

Confirms a person was actually visible on camera during the check, and flags if screen share was active instead.

How understandable your voice actually was, based on what was captured — not just whether a mic was detected.

Measures ambient noise picked up by your mic during the call — fans, traffic, other rooms — separate from your speech.

Raw dB readings for your room's quietest moments versus your actual speaking volume, so quiet or overly hot audio is caught directly.

The gap between your voice and everything else in the room — the single best predictor of how clean you'll sound to the far end.

Detects audio distortion from a mic gain set too hot, which reads as harsh or crackling on the far end.

During a two-way check, listens for echo, choppiness, or dropouts caused by your setup's echo cancellation struggling.

You're asked to read a short phrase aloud; mirrormeet compares what it actually heard against what you were supposed to say.

Clap once and mirrormeet measures how long the clap's echo takes to fade by 60 dB, with a graph of the clap dying away. The fade is read two ways, over the first 20 dB (T20) and the first 30 dB (T30), then extended to a full 60 dB. For clear speech on calls, 0.65 seconds or less is the goal.

A test sweep plays through your speaker and is recorded in the room. mirrormeet reports reverb time (RT60) and early decay (EDT) in six octave bands from 125 Hz to 4 kHz, so you can see whether bass, speech or treble is the problem.

Compares the sound that arrives in the first 50 milliseconds with what follows. A high number means speech stays crisp; a low one means the echo is blurring it.

How evenly your speaker, the room and the microphone carry low, middle and high pitches, relative to the room's own average, so peaks and dips show where the room colors the sound.

The room's size predicts where a fast, buzzy flutter echo could form between parallel walls, and the clap is checked at exactly those bounce times. Standing-wave frequencies (room modes) are worked out from the room's dimensions.

The Room Design planner estimates reverb time from the room's volume and how much sound its surfaces soak up (RT60 = 0.161 × volume ÷ absorption), live as you add furniture, rugs and drapes. A measured clap or sweep then confirms it, and a big gap means something in the room is not as described.

A room sounds different from different seats, so Room Acoustics measures at several positions and reports the average and the spread.

For a conference room, the test also runs through the room system itself in a video meeting: a helper plays the sweep through the room's loudspeakers and records what the system sends to the far end. That shows how its microphones, noise suppression and echo cancellation treat sudden sounds, and how the installed loudspeakers compare with a reference speaker. A room-only option measures just the room, with no video call and no test of the installed system.

Measured live during the test and compared against Zoom, Teams, and Meet's own published bandwidth requirements.

Round-trip latency and how much it varies — jitter is often the real cause of choppy audio even when raw speed looks fine.

The actual encoding bitrate and fps your call delivered, not just what your connection is theoretically capable of.

How long it took to actually connect and start streaming after joining — a slow join often points to a wider connection issue.

Duration, participant count, and how the session ended, so a dropped or unstable connection is caught, not just a quiet call.

Every processing stage timed and reported, so you always know exactly what's happening between your test and your results.