| // @ts-check | |
| /** | |
| * AudioWorkletProcessor that downsamples the mic to 16 kHz Float32 and hands | |
| * fixed 512-sample frames (32 ms, what Silero VAD expects) straight to the VAD | |
| * worker over a transferred MessagePort, so a busy main thread can't drop them. | |
| * | |
| * main -> worklet: { kind: "enable", value: boolean }, { kind: "port", port } | |
| */ | |
| const TARGET_RATE = 16000; | |
| const FRAME = 512; | |
| class MicCaptureProcessor extends AudioWorkletProcessor { | |
| constructor() { | |
| super(); | |
| this._step = sampleRate / TARGET_RATE; | |
| this._sum = 0; | |
| this._count = 0; | |
| this._phase = 0; | |
| this._frame = new Float32Array(FRAME); | |
| this._fill = 0; | |
| this._enabled = true; | |
| /** @type {MessagePort | null} */ | |
| this._out = null; | |
| this.port.onmessage = (e) => { | |
| if (e.data?.kind === "enable") { | |
| this._enabled = !!e.data.value; | |
| this._fill = 0; | |
| } else if (e.data?.kind === "port") { | |
| this._out = e.data.port; | |
| } | |
| }; | |
| } | |
| process(inputs) { | |
| const input = inputs[0]; | |
| const ch = input?.[0]; | |
| if (!ch || !this._enabled || !this._out) return true; | |
| for (let i = 0; i < ch.length; i++) { | |
| // Box average over one output period: a cheap low-pass that handles any | |
| // context rate (44.1 k on many Macs, 48 k elsewhere). | |
| this._sum += ch[i]; | |
| this._count += 1; | |
| this._phase += 1; | |
| if (this._phase >= this._step) { | |
| this._phase -= this._step; | |
| this._frame[this._fill++] = this._sum / this._count; | |
| this._sum = 0; | |
| this._count = 0; | |
| if (this._fill === FRAME) { | |
| const frame = this._frame; | |
| this._out.postMessage(frame, [frame.buffer]); | |
| this._frame = new Float32Array(FRAME); | |
| this._fill = 0; | |
| } | |
| } | |
| } | |
| return true; | |
| } | |
| } | |
| registerProcessor("mic-capture", MicCaptureProcessor); | |