Reduce Gesture Synth Lag and Audio Delay

Separate Tracking Lag from Audio Lag

Watch the Left and Right indicators and the live chord label while forming a stable shape. If those labels react late, investigate camera frame rate, lighting, device load and model inference. If the labels react promptly but sound arrives later, investigate the audio output path instead.

Test with a simple degree-I triad rather than a rapid song. A repeatable single gesture gives a clearer comparison than several moving controls and makes it easier to confirm which stage improved.

Use a four-pass comparison and change only one condition per pass. First, hold the I triad with the normal visual effects and current speakers. Second, keep the audio path unchanged but enable Low-performance mode. Third, keep that setting and add soft front light. Fourth, switch from Bluetooth to the built-in speaker or wired headphones. For every pass, watch whether the hand indicator, chord label and sound respond in the same order. Write down which boundary improves: camera to indicator, indicator to label or label to sound. A consistent result across several slow poses is more useful than an isolated fast response, because gesture smoothing deliberately avoids reacting to every single noisy frame.

Enable Low-Performance Mode on Slower Devices

Open Settings and enable Low-performance mode. The app reduces visual detail and the hand-tracking update rate so the device spends less time drawing effects and running inference. This can make interaction steadier when a laptop or phone drops frames under the full visual workload.

Close graphics-heavy tabs, video calls and background screen recording, then connect a laptop to power. Thermal throttling and battery-saving modes can reduce sustained camera and graphics performance even when the page initially feels responsive.

Judge Low-performance mode by stability as well as raw speed. Hold one unchanged pose for several beats and look for fewer label dropouts, then alternate I and V at a slow, even rate. A reduced update rate can make the animation look less fluid while still producing more dependable musical changes on constrained hardware. If the labels remain late after other camera applications are closed, reload once and run the same two-shape test before increasing motion speed. If only the sound remains late, leave the tracking setting alone and test the direct audio path; reducing visual work cannot remove buffering inside a Bluetooth device.

• Enable Low-performance mode in Settings.
• Close video calls and graphics-heavy tabs.
• Connect power and disable aggressive battery saving for the test.
• Restart the browser after a long, resource-heavy session.

Improve the Camera Frame Before Increasing Speed

A dark webcam may lengthen exposure and add motion blur, making each landmark estimate less stable. Use soft front lighting, avoid backlight and move far enough away to keep both wrists visible. Deliberate poses give the stabiliser a clean signal faster than rapid blurred motion.

More camera resolution is not automatically faster. A stable built-in webcam under good light can feel more responsive than a high-resolution external camera passing through a slow hub.

Follow the webcam setup checklist →

Test Wired Audio Instead of Bluetooth

Bluetooth headphones and speakers can buffer audio after the browser has already scheduled it. Compare the same gesture through wired headphones or the device speaker. If the visual chord changes immediately and only Bluetooth sounds late, hand tracking is not the primary problem.

Keep the tab unmuted and avoid routing sound through virtual audio devices during the first test. Once the direct path feels correct, add external devices one at a time so a new delay has an identifiable cause.

Fix missing or muted audio → · Read how hand tracking controls sound →