The Web Audio API PannerNode provides two primary panning models: "equalpower" (simple stereo cross-fading based on azimuth) and "HRTF" (Head-Related Transfer Function). In HRTF mode, the browser convolves the audio signal with measured human head/ear acoustic impulse response filters. It reproduces true binaural cues including Interaural Time Difference (ITD, time delay between ears) and Interaural Level Difference (ILD, head acoustic shadow frequency attenuation), allowing listeners on headphones to perceive sound height, depth, and 360-degree directionality.

The Web Audio distanceModel dictates how sound volume attenuates over distance d: "inverse" uses gain = refDistance / (refDistance + rolloffFactor * (max(d, refDistance) - refDistance)), matching authentic inverse-square physical acoustic decay in free air. "linear" applies gain = 1 - rolloffFactor * (d - refDistance) / (maxDistance - refDistance), often used in games with strict sound cutoff radii. "exponential" applies gain = (d / refDistance)^(-rolloffFactor), yielding dramatic steep falloff close to the source.

A PannerNode can focus audio in a specific direction like a megaphone or flashlight beam using coneInnerAngle, coneOuterAngle, and coneOuterGain. Sounds located within the inner angle project at full volume. Outside the outer angle, sound is attenuated to coneOuterGain. Between the inner and outer angle, volume smoothly transitions linearly, allowing simulation of actors facing away from the listener or directional speakers.

Legacy Web Audio methods like panner.setPosition(x, y, z) and listener.setOrientation(...) run on the main JavaScript thread without audio-rate scheduling, causing clicks and zipper noise during rapid movement. Modern Web Audio specifications expose positionX, positionY, positionZ as AudioParams, allowing sample-accurate automation via setValueAtTime(), linearRampToValueAtTime(), or direct connection to AudioWorklet streams for glitch-free 3D Doppler motion.