@audc/libsamplerate NPM

node-libsamplerate

ABI stable native implementation of libsamplerate as a Transform stream. Built from the latest libsamplerate code. Uses N-API, node-addon-api and cmake-js. This module has no external dependencies.

Introduction

Allows the upsampling/downsampling and/or upconverting/downconverting to/from arbitrary sample rates and to/from 16 or 32 bits per sample. Tested on Linux (x64 and arm), Windows and MacOS. This module uses the "Full Api" detailed here

Install

npm install @audc/libsamplerate --save

Requires cmake and a valid toolchain to build.

For Windows, install the Visual C++ build tools and download cmake from cmake.org. Or install Visual Studio with full c++ support.

Usage

const { SampleRate, SRC_SINC_MEDIUM_QUALITY } = require('@audc/libsamplerate');

const options = {
    // Value can be from 0 to 4 or using enum. 0 is the best quality and the slowest.
    type: SRC_SINC_MEDIUM_QUALITY,
    // Stereo
    channels: 2,
    // Sample rate of source
    fromRate: 48000,
    // bit depth of source. Valid values: 16 or 32
    fromDepth: 16,
    // Desired sample rate
    toRate: 44100,
    // Desired bit depth. Valid values: 16 or 32
    toDepth: 16
}
const resample = new SampleRate(options);

Input audio data should be signed integers (e.g. S16_LE or S32_LE). Output will also be signed integers. Floating point input/output is not yet supported. Input should be from a readable stream, output should be to a writable stream: e.g.

const fs = require('fs');
const rs = fs.createReadStream('input.pcm');
const ws = fs.createWriteStream('output.pcm');

rs.pipe(resample).pipe(ws);

NOTE: if reading from a WAV file, start the read at 44 bytes to avoid the wav header ({start:44} for fs.createReadStream). If recording from arecord, sox or similar, use format raw.

Altering playback speed

This is possible by modifying the converter ratio, which can be done on the fly using the setRatio method. e.g. resample.setRatio(0.5). Depending upon timing, this will most likely take effect at the next buffer load. If fine grained modifications to playback speed are required, then a small value for highWaterMark on the input stream will be needed.