AV Sync Test — Lip-Sync Calibration

An AV sync test measures lip-sync error by flashing a visual target and scheduling an audio click, then letting you dial the offset between them. Both events are generated in your browser; the click is scheduled on the Web Audio clock, and the offset slider covers −300 to +300 ms.

How to use it

1. Press Start AV Sync Test, or tap Space. The status badge changes to Calibrating Live and the five-step cadence begins.
2. Listen to the first four ticks. Steps 1 to 4 are preparatory pips that establish the tempo. Leave Prep Audio Pips on while you find your bearings, then switch it off once you can anticipate the beat. Choose 100, 120 or 150 BPM; 120 BPM spaces the steps 500 ms apart.
3. Watch step 5, the SYNC step. The reticle flashes on the fifth step at the same moment the primary click is scheduled. If the click lands early the picture is late; if it lands late, the audio path is slow.
4. Drag the offset slider until they coincide. The range is −300 to +300 ms in 5 ms steps, with nudge buttons for ±1, ±5 and ±10 ms and a reset to 0. Negative values play the click earlier than the flash; positive values play it later.
5. Read the analysis card. It classifies the current offset against ITU-R BT.1359-1 as Broadcast, Acceptable, Detectable or Excessive, and suggests a correction.
6. Apply the number to your hardware. Enter it as lip-sync or audio delay in your AVR or soundbar, or press a preset to compare Bluetooth and display-lag cases.

What the test actually measures

The test compares two chains that should agree: the display path that draws the flash and the audio path that plays the click. The click is scheduled ahead of time on the AudioContext clock, so its timing does not depend on JavaScript timers. The flash is drawn on the next animation frame, so it inherits the whole video chain: compositor, TV processing, panel response and any motion smoothing in between. The offset you dial in is the difference between the two, which is exactly the quantity a lip-sync control corrects.

The display sets a floor on resolution: a 60 Hz screen redraws every 16.7 ms, so one flash carries a 17 ms quantization step; a 120 Hz panel halves that to 8.3 ms. The ±15 ms broadcast target is at the edge of what one flash can show, so judge over several cycles rather than trusting a single hit.

The ITU-R BT.1359-1 window

ITU-R BT.1359-1, “Relative timing of sound and vision for broadcasting”, reports average detectability thresholds of about +45 ms when sound leads the picture and −125 ms when sound lags it, with acceptability thresholds of about +90 ms and −185 ms. The asymmetry is physical: sound travels at 343 m/s, so a person 15 m away is naturally about 44 ms late, while light arrives effectively instantly. Late sound feels plausible; early sound does not.

Timing relationship	Detectability threshold	Acceptability threshold	What viewers notice
Sound leads picture (advanced)	+45 ms	+90 ms	Speech reads as dubbed and unnatural, because sound never precedes light in nature
Sound lags picture (delayed)	−125 ms	−185 ms	Tolerated far longer, matching the natural delay of distant sound
Broadcast reference	±15 ms	Not applicable	Studio-grade target; imperceptible in dialogue

The tool’s own classification follows the same numbers: within ±15 ms it reports Broadcast grade, up to 45 ms of lead or 125 ms of lag it reports Acceptable, then Detectable up to 90 ms or 185 ms, and Excessive beyond that.

Applying the correction

• AV receivers and soundbars expose a lip-sync, audio delay or AV sync control in milliseconds. If the test says audio lags, add delay to the video side or reduce audio delay, depending on which direction your device’s control works.
• Smart TVs often lose 60 to 180 ms to motion smoothing, noise reduction and upscaling. Game or PC mode bypasses most of that processing and can remove the error at the source.
• Desktop players usually carry an audio track sync control. VLC uses J and K to nudge the audio track, and many players let you type an exact millisecond value.
• Bluetooth audio buffers packets, so the delay depends on the codec and the operating system stack. Low-latency modes such as aptX-LL or LC3 cut the number substantially; a cable removes it.
• Measure with the output you actually watch with. A setting that lands perfectly on the TV speakers can be 150 ms off over Bluetooth headphones, because each output path buffers differently.

Accuracy, limits and troubleshooting

• Judge from your normal seat. The flash reaches your eyes almost instantly, but the click still takes about 3 ms per metre to reach your ears. At 3 m from the screen that is roughly 9 ms of real delay the tool cannot know about, so do not calibrate from 30 cm away and then sit on the sofa.
• Keep the tab in the foreground. Browsers throttle animation frames in background tabs, which degrades flash timing; the audio click stays accurate, so a backgrounded tab reports a false offset.
• One device, one measurement. The offset is a property of the whole chain, so a different browser, a different cable or a different soundbar input can produce a different number.
• Relative, not absolute. The test measures the audio-to-video difference. For the total latency of a path, including Bluetooth buffering, use the Audio Latency Test.
• Wireless measurement is noisier. Bluetooth and Wi-Fi audio can drop or re-buffer during a run; repeat the test and look for a consistent reading before you change a setting.
• Do not chase a number you cannot hear. If dialogue looks and sounds right, leave the settings alone; the ±15 ms broadcast window exists because smaller errors are not perceptible in normal viewing. A speaker far off-axis or a strongly reflective room can make an accurately synced system feel late, so check the room before blaming the timing chain.

Related tools

• Audio Latency Test: measure absolute delay for Bluetooth, soundbars and wired paths.
• Speaker Test: verify channel assignment, polarity and centre imaging before judging sync.
• Headphone Test: check driver balance and extension on the output you watch with.
• Surround Sound Test: confirm discrete channel routing in a 5.1 or 7.1 system.