@picovoice/rhino-web NPM

Rhino Binding for Web

Rhino Speech-to-Intent engine

Made in Vancouver, Canada by Picovoice

Rhino is Picovoice's Speech-to-Intent engine. It directly infers intent from spoken commands within a given context of interest, in real-time. For example, given a spoken command:

Can I have a small double-shot espresso?

Rhino infers that the user would like to order a drink and emits the following inference result:

{
  "isUnderstood": "true",
  "intent": "orderBeverage",
  "slots": {
    "beverage": "espresso",
    "size": "small",
    "numberOfShots": "2"
  }
}

Rhino is:

using deep neural networks trained in real-world environments.
compact and computationally-efficient, making it perfect for IoT.
self-service. Developers and designers can train custom models using Picovoice Console.

Compatibility

Chrome / Edge
Firefox
Safari

Restrictions

IndexedDB is required to use Rhino in a worker thread. Browsers without IndexedDB support (i.e. Firefox Incognito Mode) should use Rhino in the main thread.

Installation

Package

Using Yarn:

yarn add @picovoice/rhino-web

or using npm:

npm install --save @picovoice/rhino-web

AccessKey

Rhino requires a valid Picovoice AccessKey at initialization. AccessKey acts as your credentials when using Rhino SDKs. You can get your AccessKey for free. Make sure to keep your AccessKey secret. Signup or Login to Picovoice Console to get your AccessKey.

Usage

There are two methods to initialize Rhino:

Public Directory

NOTE: Due to modern browser limitations of using a file URL, this method does not work if used without hosting a server.

This method fetches the model file from the public directory and feeds it to Rhino. Copy the model file into the public directory:

cp ${RHINO_MODEL_FILE} ${PATH_TO_PUBLIC_DIRECTORY}

The same procedure can be used for the Rhino context (.rhn) files.

Base64

NOTE: This method works without hosting a server, but increases the size of the model file roughly by 33%.

This method uses a base64 string of the model file and feeds it to Rhino. Use the built-in script pvbase64 to base64 your model file:

npx pvbase64 -i ${RHINO_MODEL_FILE} -o ${OUTPUT_DIRECTORY}/${MODEL_NAME}.js

The output will be a js file which you can import into any file of your project. For detailed information about pvbase64, run:

npx pvbase64 -h

The same procedure can be used for the Rhino context (.rhn) files.

Rhino Model

Rhino saves and caches your model (.pv) and context (.rhn) files in the IndexedDB to be used by Web Assembly. Use a different customWritePath variable to hold multiple model values and set the forceWrite value to true to force an overwrite of the model file. If the model (.pv) or context (.rhn) files change, version should be incremented to force the cached model to be updated. Either base64 or publicPath must be set to instantiate Rhino. If both are set, Rhino will use the base64 parameter.

// Context (.rhn)
const rhinoContext = {
  publicPath: ${CONTEXT_RELATIVE_PATH},
  // or
  base64: ${CONTEXT_BASE64_STRING},

  // Optionals
  customWritePath: 'custom_context',
  forceWrite: true,
  version: 1,
  sensitivity: 0.5,
}

// Model (.pv)
const rhinoModel = {
  publicPath: ${MODEL_RELATIVE_PATH},
  // or
  base64: ${MODEL_BASE64_STRING},

  // Optionals
  customWritePath: 'custom_model',
  forceWrite: true,
  version: 1,
}

Additional engine options are provided via the options parameter. Set processErrorCallback to handle errors if an error occurs while processing audio. Use endpointDurationSec and requireEndpoint to control the engine's endpointing behaviour. An endpoint is a chunk of silence at the end of an utterance that marks the end of spoken command.

// Optional. These are the default values
const options = {
  endpointDurationSec: 1.0,
  requireEndpoint: true,
  processErrorCallback: (error) => {},
}

Initialize Rhino

Create a inferenceCallback function to get the results from the engine:

function inferenceCallback(inference) {
  if (inference.isFinalized) {
    if (inference.isUnderstood) {
      console.log(inference.intent)
      console.log(inference.slots)
    }
  }
}

Create an options object and add a processErrorCallback function if you would like to catch errors:

function processErrorCallback(error: string) {
...
}
options.processErrorCallback = processErrorCallback;

Initialize an instance of Rhino in the main thread:

const handle = await Rhino.create(
  ${ACCESS_KEY},
  rhinoContext,
  inferenceCallback,
  rhinoModel,
  options // optional options
);

Or initialize an instance of Rhino in a worker thread:

const handle = await RhinoWorker.create(
  ${ACCESS_KEY},
  rhinoContext,
  inferenceCallback,
  rhinoModel,
  options // optional options
);

Process Audio Frames

The result is received from inferenceCallback as defined above.

function getAudioData(): Int16Array {
... // function to get audio data
  return new Int16Array();
}
for (; ;) {
  await handle.process(getAudioData());
  // break on some condition
}

Clean Up

Clean up used resources by Rhino or RhinoWorker:

await handle.release();

Terminate

Terminate RhinoWorker instance:

await handle.terminate();

Contexts

Create custom contexts using the Picovoice Console. Train and download a Rhino context file (.rhn) for the target platform Web (WASM). This model file can be used directly with publicPath, but, if base64 is preferable, convert the .rhn file to a base64 JavaScript variable using the built-in pvbase64 script:

npx pvbase64 -i ${CONTEXT_FILE}.rhn -o ${CONTEXT_BASE64}.js -n ${CONTEXT_BASE64_VAR_NAME}

Similar to the model file (.pv), context files (.rhn) are saved in IndexedDB to be used by Web Assembly. Either base64 or publicPath must be set for the context to instantiate Rhino. If both are set, Rhino will use the base64 model.

const contextModel = {
  publicPath: "${CONTEXT_RELATIVE_PATH}",
  // or
  base64: "${CONTEXT_BASE64_STRING}",
}

Switching Languages

In order to make inferences in different language you need to use the corresponding model file (.pv). The model files for all supported languages are available here.