Spokestack

Spokestack: give your React Native app a voice interface!

Description

React Native plugin for adding voice using Spokestack. This includes speech recognition, wakeword, and natural language understanding, as well as synthesizing text to speech using Spokestack voices.

Programming language: TypeScript

License: Apache License 2.0

Tags: Audio / Video     Dev Tools     React     Reactjs     Library    

README

[React Native Spokestack](./example/react-native-spokestack.png)

React Native plugin for adding voice using Spokestack. This includes speech recognition, wakeword, and natural language understanding, as well as synthesizing text to speech using Spokestack voices.

Requirements

• React Native: 0.60.0+
• Android: Android SDK 21+
• iOS: iOS 13+

Installation

Using npm:

npm install --save react-native-spokestack

or using yarn:

yarn add react-native-spokestack

Then follow the instructions for each platform to link react-native-spokestack to your project:

iOS installation

iOS details

Set deployment target

First, open XCode and go to Project -> Info to set the iOS Deployment target to 13.0 or higher.

Also, set deployment to 13.0 under Target -> General -> Deployment Info.

Remove invalid library search path

When Flipper was introduced to React Native, some library search paths were set for Swift. There has been a longstanding issue with the default search paths in React Native projects because a search path was added for swift 5.0 which prevented any other React Native libraries from using APIs only available in Swift 5.2 or later. Spokestack-iOS, a dependency of react-native-spokestack makes use of these APIs and XCode will fail to build.

Fortunately, the fix is fairly simple. Go to your target -> Build Settings and search for "Library Search Paths".

Remove "\"$(TOOLCHAIN_DIR)/usr/lib/swift-5.0/$(PLATFORM_NAME)\"" from the list.

Edit Podfile

Before running pod install, make sure to make the following edits.

react-native-spokestack makes use of relatively new APIs only available in iOS 13+. Set the deployment target to iOS 13 at the top of your Podfile:

platform :ios, '13.0'

We also need to use use_frameworks! in our Podfile in order to support dependencies written in Swift.

target 'SpokestackExample' do
  use_frameworks!
  #...

For now, use_frameworks! does not work with Flipper, so we also need to disable Flipper. Remove any Flipper-related lines in your Podfile. In React Native 0.63.2+, they look like this:

  # X Remove or comment out these lines X
  # use_flipper!
  # post_install do |installer|
  #   flipper_post_install(installer)
  # end
  # XX

Bug in React Native 0.64.0 (should be fixed in 0.64.1)

React Native 0.64.0 broke any projects using use_frameworks! in their Podfiles.

For more info on this bug, see https://github.com/facebook/react-native/issues/31149.

To workaround this issue, add the following to your Podfile:

# Moves 'Generate Specs' build_phase to be first for FBReactNativeSpec
post_install do |installer|
  installer.pods_project.targets.each do |target|
    if (target.name&.eql?('FBReactNativeSpec'))
      target.build_phases.each do |build_phase|
        if (build_phase.respond_to?(:name) && build_phase.name.eql?('[CP-User] Generate Specs'))
          target.build_phases.move(build_phase, 0)
        end
      end
    end
  end
end

pod install

Remove your existing Podfile.lock and Pods folder to ensure no conflicts, then install the pods:

$ npx pod-install

Edit Info.plist

Add the following to your Info.plist to enable permissions.

<key>NSMicrophoneUsageDescription</key>
<string>This app uses the microphone to hear voice commands</string>
<key>NSSpeechRecognitionUsageDescription</key>
<string>This app uses speech recognition to process voice commands</string>

Remove Flipper

While Flipper works on fixing their pod for use_frameworks!, we must disable Flipper. We already removed the Flipper dependencies from Pods above, but there remains some code in the AppDelegate.m that imports Flipper. There are two ways to fix this.

1. You can disable Flipper imports without removing any code from the AppDelegate. To do this, open your xcworkspace file in XCode. Go to your target, then Build Settings, search for "C Flags", remove -DFB_SONARKIT_ENABLED=1 from flags.
2. Remove all Flipper-related code from your AppDelegate.m.

In our example app, we've done option 1 and left in the Flipper code in case they get it working in the future and we can add it back.

Edit AppDelegate.m

Add AVFoundation to imports

#import <AVFoundation/AVFoundation.h>

AudioSession category

Set the AudioSession category. There are several configurations that work.

The following is a suggestion that should fit most use cases:

- (BOOL)application:(UIApplication *)application didFinishLaunchingWithOptions:(NSDictionary *)launchOptions
{
  AVAudioSession *session = [AVAudioSession sharedInstance];
  [session setCategory:AVAudioSessionCategoryPlayAndRecord
     mode:AVAudioSessionModeDefault
  options:AVAudioSessionCategoryOptionDefaultToSpeaker | AVAudioSessionCategoryOptionAllowAirPlay | AVAudioSessionCategoryOptionAllowBluetoothA2DP | AVAudioSessionCategoryOptionAllowBluetooth
    error:nil];
  [session setActive:YES error:nil];

  // ...

Android installation

Android details

ASR Support

The example usage uses the system-provided ASRs (AndroidSpeechRecognizer and AppleSpeechRecognizer). However, AndroidSpeechRecognizer is not available on 100% of devices. If your app supports a device that doesn't have built-in speech recognition, use Spokestack ASR instead by setting the profile to a Spokestack profile using the profile prop.

See our ASR documentation for more information.

Edit root build.gradle (not app/build.gradle)

// ...
  ext {
    // Set the minimum SDK to 24.
    // React Native 0.64+ sets version 21.
    // If you prefer to leave the minimum SDK at 21,
    // another option is to set this to 21, but
    // also set android.enableDexingArtifactTransform=false
    // in your top-level gradle.properties.
    // See https://github.com/google/ExoPlayer/issues/6801#issuecomment-574089568
    minSdkVersion = 24
// ...
  dependencies {
    // Minimium gradle is 3.0.1+
    // The latest React Native already has this
    classpath("com.android.tools.build:gradle:4.2.1")

Edit AndroidManifest.xml

Add the necessary permissions to your AndroidManifest.xml. The first permission is often there already. The second is needed for using the microphone.

    <!-- For TTS -->
    <uses-permission android:name="android.permission.INTERNET" />
    <!-- For wakeword and ASR -->
    <uses-permission android:name="android.permission.RECORD_AUDIO" />
    <!-- For ensuring no downloads happen over cellular, unless forced -->
    <uses-permission android:name="android.permission.ACCESS_NETWORK_STATE" />

Request RECORD_AUDIO permission

The RECORD_AUDIO permission is special in that it must be both listed in the AndroidManifest.xml as well as requested at runtime. There are a couple ways to handle this (react-native-spokestack does not do this for you):

1. Recommended Add a screen to your onboarding that explains the need for the permissions used on each platform (RECORD_AUDIO on Android and Speech Recognition on iOS). Have a look at react-native-permissions to handle permissions in a more robust way.
2. Request the permissions only when needed, such as when a user taps on a "listen" button. Avoid asking for permission with no context or without explaining why it is needed. In other words, we do not recommend asking for permission on app launch.

While iOS will bring up permissions dialogs automatically for any permissions needed, you must do this manually in Android.

React Native already provides a module for this. See React Native's PermissionsAndroid for more info.

Usage

Get started using Spokestack, or check out our in-depth tutorials on ASR, NLU, and TTS. Also be sure to take a look at the Cookbook for quick solutions to common problems.

A working example app is included in this repo in the example/ folder.

import Spokestack from 'react-native-spokestack'
import { View, Button, Text } from 'react-native'

function App() {
  const [listening, setListening] = useState(false)

  const onActivate = () => setListening(true)
  const onDeactivate = () => setListening(false)
  const onRecognize = ({ transcript }) => console.log(transcript)

  useEffect(() => {
    Spokestack.addEventListener('activate', onActivate)
    Spokestack.addEventListener('deactivate', onDeactivate)
    Spokestack.addEventListener('recognize', onRecognize)
    Spokestack.initialize(
      process.env.SPOKESTACK_CLIENT_ID,
      process.env.SPOKESTACK_CLIENT_SECRET
    )
      // This example starts the Spokestack pipeline immediately,
      // but it could be delayed until after onboarding or other
      // conditions have been met.
      .then(Spokestack.start)

    return () => {
      Spokestack.removeAllListeners()
    }
  }, [])

  return (
    <View>
      <Button onClick={() => Spokestack.activate()} title="Listen" />
      <Text>{listening ? 'Listening...' : 'Idle'}</Text>
    </View>
  )
}

Including model files in your app bundle

To include model files locally in your app (rather than downloading them from a CDN), you also need to add the necessary extensions so the files can be included by Babel. To do this, edit your metro.config.js.

const defaults = require('metro-config/src/defaults/defaults')

module.exports = {
  resolver: {
    assetExts: defaults.assetExts.concat(['tflite', 'txt', 'sjson'])
  }
}

Then include model files using source objects:

Spokestack.initialize(clientId, clientSecret, {
  wakeword: {
    filter: require('./filter.tflite'),
    detect: require('./detect.tflite'),
    encode: require('./encode.tflite')
  },
  nlu: {
    model: require('./nlu.tflite'),
    vocab: require('./vocab.txt'),
    // Be sure not to use "json" here.
    // We use a different extension (.sjson) so that the file is not
    // immediately parsed as json and instead
    // passes a require source object to Spokestack.
    // The special extension is only necessary for local files.
    metadata: require('./metadata.sjson')
  }
})

This is not required. Pass remote URLs to the same config options and the files will be downloaded and cached when first calling initialize.

Contributing

See the [contributing guide](CONTRIBUTING.md) to learn how to contribute to the repository and the development workflow.

---

API Documentation

initialize

▸ initialize(clientId, clientSecret, config?): Promise<void>

Initialize the speech pipeline; required for all other methods.

The first 2 args are your Spokestack credentials available for free from https://spokestack.io. Avoid hardcoding these in your app. There are several ways to include environment variables in your code.

Using process.env: https://babeljs.io/docs/en/babel-plugin-transform-inline-environment-variables/

Using a local .env file ignored by git: https://github.com/goatandsheep/react-native-dotenv https://github.com/luggit/react-native-config

See SpokestackConfig for all available options.

example

import Spokestack from 'react-native-spokestack'

// ...

await Spokestack.initialize(process.env.CLIENT_ID, process.env.CLIENT_SECRET, {
  pipeline: {
    profile: Spokestack.PipelineProfile.PTT_NATIVE_ASR
  }
})

Parameters

Name	Type
clientId	string
clientSecret	string
config?	[SpokestackConfig](#SpokestackConfig)

Returns

Promise<void>

Defined in

index.ts:64

destroy

▸ destroy(): Promise<void>

Destroys the speech pipeline, removes all listeners, and frees up all resources. This can be called before re-initializing the pipeline. A good place to call this is in componentWillUnmount.

example

componentWillUnmount() {
  Spokestack.destroy()
}

Returns

Promise<void>

Defined in

index.ts:81

start

▸ start(): Promise<void>

Start the speech pipeline. The speech pipeline starts in the deactivate state.

example

import Spokestack from 'react-native-spokestack`

// ...

Spokestack.initialize(process.env.CLIENT_ID, process.env.CLIENT_SECRET)
  .then(Spokestack.start)

Returns

Promise<void>

Defined in

index.ts:96

stop

▸ stop(): Promise<void>

Stop the speech pipeline. This effectively stops ASR, VAD, and wakeword.

example

import Spokestack from 'react-native-spokestack`

// ...

await Spokestack.stop()

Returns

Promise<void>

Defined in

index.ts:110

activate

▸ activate(): Promise<void>

Manually activate the speech pipeline. This is necessary when using a PTT profile. VAD profiles can also activate ASR without the need to call this method.

example

import Spokestack from 'react-native-spokestack`

// ...

<Button title="Listen" onClick={() => Spokestack.activate()} />

Returns

Promise<void>

Defined in

index.ts:126

deactivate

▸ deactivate(): Promise<void>

Deactivate the speech pipeline. If the profile includes wakeword, the pipeline will go back to listening for the wakeword. If VAD is active, the pipeline can reactivate without calling activate().

example

import Spokestack from 'react-native-spokestack`

// ...

<Button title="Stop listening" onClick={() => Spokestack.deactivate()} />

Returns

Promise<void>

Defined in

index.ts:142

synthesize

▸ synthesize(input, format?, voice?): Promise<string>

Synthesize some text into speech Returns Promise<string> with the string being the URL for a playable mpeg.

There is currently only one free voice available ("demo-male"). The voice can be changed if you have created a custom voice using a Spokestack Maker account. See https://spokestack.io/pricing#maker.

example

const url = await Spokestack.synthesize('Hello world')
play(url)

Parameters

Name	Type
input	string
format?	[TTSFormat](#TTSFormat)
voice?	string

Returns

Promise<string>

Defined in

index.ts:158

speak

▸ speak(input, format?, voice?): Promise<void>

Synthesize some text into speech and then immediately play the audio through the default audio system. Audio session handling can get very complex and we recommend using a RN library focused on audio for anything more than very simple playback.

There is currently only one free voice available ("demo-male").

example

await Spokestack.speak('Hello world')

Parameters

Name	Type
input	string
format?	[TTSFormat](#TTSFormat)
voice?	string

Returns

Promise<void>

Defined in

index.ts:174

classify

▸ classify(utterance): Promise<SpokestackNLUResult>

Classify the utterance using the intent/slot Natural Language Understanding model passed to Spokestack.initialize(). See https://www.spokestack.io/docs/concepts/nlu for more info.

example

const result = await Spokestack.classify('hello')

// Here's what the result might look like,
// depending on the NLU model
console.log(result.intent) // launch

Parameters

Name	Type
utterance	string

Returns

Promise<SpokestackNLUResult>

Defined in

index.ts:190

isInitialized

▸ isInitialized(): Promise<boolean>

Returns whether Spokestack has been initialized

example

console.log(`isInitialized: ${await Spokestack.isInitialized()}`)

Returns

Promise<boolean>

Defined in

index.ts:199

isStarted

▸ isStarted(): Promise<boolean>

Returns whether the speech pipeline has been started

example

console.log(`isStarted: ${await Spokestack.isStarted()}`)

Returns

Promise<boolean>

Defined in

index.ts:208

isActivated

▸ isActivated(): Promise<boolean>

Returns whether the speech pipeline is currently activated

example

console.log(`isActivated: ${await Spokestack.isActivated()}`)

Returns

Promise<boolean>

Defined in

index.ts:217

SpokestackNLUResult

confidence

• confidence: number

A number from 0 to 1 representing the NLU model's confidence in the intent it recognized, where 1 represents absolute confidence.

Defined in

types.ts:115

intent

• intent: string

The intent based on the match provided by the NLU model

Defined in

types.ts:113

slots

• slots: SpokestackNLUSlots

Data associated with the intent, provided by the NLU model

Defined in

types.ts:117

Interface: SpokestackNLUSlots

Indexable

▪ [key: string]: SpokestackNLUSlot

SpokestackNLUSlot

rawValue

• rawValue: string

The original string value of the slot recognized in the user utterance

Defined in

types.ts:104

type

• type: string

The slot's type, as defined in the model metadata

Defined in

types.ts:100

value

• value: any

The parsed (typed) value of the slot recognized in the user utterance

Defined in

types.ts:102

addEventListener

▸ addEventListener(eventType, listener, context?): EmitterSubscription

Bind to any event emitted by the native libraries See Events for a list of all available events.

example

useEffect(() => {
  const listener = Spokestack.addEventListener('recognize', onRecognize)
  // Unsubsribe by calling remove when components are unmounted
  return () => {
    listener.remove()
  }
}, [])

Parameters

Name	Type	Description
eventType	string	name of the event for which we are registering listener
listener	(event: any) => void	the listener function
context?	Object	context of the listener

Returns

EmitterSubscription

Defined in

index.ts:237

removeEventListener

▸ removeEventListener(eventType, listener): void

Remove an event listener

example

Spokestack.removeEventListener('recognize', onRecognize)

Parameters

Name	Type	Description
eventType	string	Name of the event to emit
listener	(...args: any[]) => any	Function to invoke when the specified event is emitted

Returns

void

Defined in

index.ts:253

removeAllListeners

▸ removeAllListeners(): void

Remove any existing listeners

example

Spokestack.removeAllListeners()

Returns

void

Defined in

index.ts:265

TTSFormat

Three formats are supported when using Spokestack TTS. Raw text, SSML, and Speech Markdown. See https://www.speechmarkdown.org/ if unfamiliar with Speech Markdown. IPA is expected when using SSML or Speech Markdown.

• SPEECHMARKDOWN = 2

Defined in

types.ts:73

• SSML = 1

Defined in

types.ts:72

• TEXT = 0

Defined in

types.ts:71

---

Events

Use addEventListener(), removeEventListener(), and removeAllListeners() to add and remove events handlers. All events are available in both iOS and Android.

Name	Data	Description
recognize	{ transcript: string }	Fired whenever speech recognition completes successfully.
partial_recognize	{ transcript: string }	Fired whenever the transcript changes during speech recognition.
start	null	Fired when the speech pipeline starts (which begins listening for wakeword or starts VAD).
stop	null	Fired when the speech pipeline stops.
activate	null	Fired when the speech pipeline activates, either through the VAD, wakeword, or when calling .activate().
deactivate	null	Fired when the speech pipeline deactivates.
play	{ playing: boolean }	Fired when TTS playback starts and stops. See the speak() function.
timeout	null	Fired when an active pipeline times out due to lack of recognition.
trace	{ message: string }	Fired for trace messages. Verbosity is determined by the traceLevel option.
error	{ error: string }	Fired when there's an error in Spokestack.

When an error event is triggered, any existing promises are rejected.

---

SpokestackConfig

These are the configuration options that can be passed to Spokestack.initialize(_, _, spokestackConfig). No options in SpokestackConfig are required.

SpokestackConfig has the following structure:

interface SpokestackConfig {
  /**
   * This option is only used when remote URLs are passed to fields such as `wakeword.filter`.
   *
   * Set this to true to allow downloading models over cellular.
   * Note that `Spokestack.initialize()` will still reject the promise if
   * models need to be downloaded but there is no network at all.
   *
   * Ideally, the app will include network handling itself and
   * inform the user about file downloads.
   *
   * Default: false
   */
  allowCellularDownloads?: boolean
  /**
   * Wakeword, Keyword, and NLU model files are cached internally.
   * Set this to true whenever a model is changed
   * during development to refresh the internal model cache.
   *
   * This affects models passed with `require()` as well
   * as models downloaded from remote URLs.
   *
   * Default: true in dev mode, false otherwise
   *
   * **Important:** By default, apps in production will
   * cache models to avoid downloading them every time
   * the app is launched. The side-effect of this optimization
   * is that if models change on the CDN, apps will
   * not pick up those changes–unless the app were reinstalled.
   * We think this is a fair trade-off, but set this to `true`
   * if you prefer to download the models every time the app
   * is launched.
   */
  refreshModels?: boolean
  /**
   * This controls the log level for the underlying native
   * iOS and Android libraries.
   * Also add a `"trace"` event listener to get trace events.
   * See the TraceLevel enum for values.
   */
  traceLevel?: TraceLevel
  /**
   * Most of these options are advanced aside from "profile"
   */
  pipeline?: PipelineConfig
  /** Only needed if using Spokestack.classify */
  nlu?: NLUConfig
  /**
   * Only required for wakeword
   * Most options are advanced aside from
   * filter, encode, and decode for specifying config files.
   */
  wakeword?: WakewordConfig
  /**
   * Only required for the keyword recognizer
   * Most options are advanced aside from
   * filter, encode, decode, metadata, and classes.
   */
  keyword?: KeywordConfig
}

TraceLevel

How much logging to show A lower number means more logs.

• DEBUG = 10

Defined in

types.ts:58

• INFO = 30

Defined in

types.ts:60

• NONE = 100

Defined in

types.ts:61

• PERF = 20

Defined in

types.ts:59

PipelineProfile

Pipeline profiles set up the speech pipeline based on your needs

• PTT_NATIVE_ASR = 2

Apple/Android Automatic Speech Recogntion is on when the speech pipeline is active. This is likely the more common profile when not using wakeword.

Defined in

types.ts:24

• PTT_SPOKESTACK_ASR = 5

Spokestack Automatic Speech Recogntion is on when the speech pipeline is active. This is likely the more common profile when not using wakeword, but Spokestack ASR is preferred.

Defined in

types.ts:42

• TFLITE_WAKEWORD_KEYWORD = 6

VAD-sensitive TFLite wake word activates TFLite keyword recognizer

Defined in

types.ts:46

• TFLITE_WAKEWORD_NATIVE_ASR = 0

Set up wakeword and use local Apple/Android ASR. Note that wakeword.filter, wakeword.encode, and wakeword.detect are required if any wakeword profile is used.

Defined in

types.ts:12

• TFLITE_WAKEWORD_SPOKESTACK_ASR = 3

Set up wakeword and use remote Spokestack ASR. Note that wakeword.filter, wakeword.encode, and wakeword.detect are required if any wakeword profile is used.

Defined in

types.ts:30

• VAD_KEYWORD_ASR = 7

VAD-triggered TFLite Keyword Recognizer

Defined in

types.ts:50

• VAD_NATIVE_ASR = 1

Apple/Android Automatic Speech Recognition is on when Voice Active Detection triggers it.

Defined in

types.ts:17

• VAD_SPOKESTACK_ASR = 4

Spokestack Automatic Speech Recognition is on when Voice Active Detection triggers it.

Defined in

types.ts:35

Pipeline Config

agcCompressionGainDb

• Optional agcCompressionGainDb: number

advanced

Android-only for AcousticGainControl

Target peak audio level, in -dB, to maintain a peak of -9dB, configure a value of 9

Defined in

types.ts:192

agcTargetLevelDbfs

• Optional agcTargetLevelDbfs: number

advanced

Android-only for AcousticGainControl

Dynamic range compression rate, in dBFS

Defined in

types.ts:200

ansPolicy

• Optional ansPolicy: "aggressive" | "very-aggressive" | "mild" | "medium"

advanced

Android-only for AcousticNoiseSuppressor

Noise policy

Defined in

types.ts:183

bufferWidth

• Optional bufferWidth: number

advanced

Buffer width, used with frameWidth to determine the buffer size

Defined in

types.ts:155

frameWidth

• Optional frameWidth: number

advanced

Speech frame width, in ms

Defined in

types.ts:149

profile

• Optional profile: PipelineProfile

Profiles are collections of common configurations for Pipeline stages.

If no profile is set explicitly, Spokestack determines a sensible default profile based on the config passed to Spokestack.initialize():

If wakeword config files are set (and keyword config is not), the default will be set to TFLITE_WAKEWORD_NATIVE_ASR.

If keyword config files are set (and wakeword config is not), the default will be set to VAD_KEYWORD_ASR.

If both wakeword and keyword config files are set, the default will be set to TFLITE_WAKEWORD_KEYWORD.

Otherwise, the default is PTT_NATIVE_ASR.

Defined in

types.ts:139

sampleRate

• Optional sampleRate: number

Audio sampling rate, in Hz

Defined in

types.ts:143

vadFallDelay

• Optional vadFallDelay: number

advanced

Falling-edge detection run length, in ms; this value determines how many negative samples must be received to flip the detector to negative

Defined in

types.ts:166

vadMode

• Optional vadMode: "quality" | "low-bitrate" | "aggressive" | "very-aggressive"

Voice activity detector mode

Defined in

types.ts:159

vadRiseDelay

• Optional vadRiseDelay: number

advanced

Android-only

Rising-edge detection run length, in ms; this value determines how many positive samples must be received to flip the detector to positive

Defined in

types.ts:175

NLU Config

metadata

• metadata: string | number

The JSON file for NLU metadata. If specified, model and vocab are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. metadata: require('./metadata.sjson')).

IMPORTANT: a special extension is used for local metadata JSON files (.sjson) when using require or import so the file is not parsed when included but instead imported as a source object. This makes it so the file is read and parsed by the underlying native libraries instead.

Defined in

types.ts:223

model

• model: string | number

The NLU Tensorflow-Lite model. If specified, metadata and vocab are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. model: require('./nlu.tflite'))

Defined in

types.ts:211

vocab

• vocab: string | number

A txt file containing the NLU vocabulary. If specified, model and metadata are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. vocab: require('./vocab.txt'))

Defined in

types.ts:231

inputLength

• Optional inputLength: number

Defined in

types.ts:244

Wakeword Config

detect

• detect: string | number

The "detect" Tensorflow-Lite model. If specified, filter and encode are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. detect: require('./detect.tflite'))

The encode model is used to perform each autoregressive step over the mel frames; its inputs should be shaped [mel-length, mel-width], and its outputs [encode-width], with an additional state input/output shaped [state-width]

Defined in

types.ts:272

encode

• encode: string | number

The "encode" Tensorflow-Lite model. If specified, filter and detect are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. encode: require('./encode.tflite'))

Its inputs should be shaped [encode-length, encode-width], and its outputs

Defined in

types.ts:283

filter

• filter: string | number

The "filter" Tensorflow-Lite model. If specified, detect and encode are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. filter: require('./filter.tflite'))

The filter model is used to calculate a mel spectrogram frame from the linear STFT; its inputs should be shaped [fft-width], and its outputs [mel-width]

Defined in

types.ts:260

activeMax

• Optional activeMax: number

advanced

The maximum length of an activation, in milliseconds, used to time out the activation

Defined in

types.ts:373

activeMin

• Optional activeMin: number

advanced

The minimum length of an activation, in milliseconds, used to ignore a VAD deactivation after the wakeword

Defined in

types.ts:366

requestTimeout

• Optional requestTimeout: number

iOS-only

Length of time to allow an Apple ASR request to run, in milliseconds. Apple has an undocumented limit of 60000ms per request.

Defined in

types.ts:380

rmsAlpha

• Optional rmsAlpha: number

advanced Android-only

The Exponentially-Weighted Moving Average (EWMA) update rate for the current RMS signal energy (0 for no RMS normalization)

Defined in

types.ts:397

rmsTarget

• Optional rmsTarget: number

advanced Android-only

The desired linear Root Mean Squared (RMS) signal energy, which is used for signal normalization and should be tuned to the RMS target used during training

Defined in

types.ts:389

wakewords

• Optional wakewords: string | string[]

iOS-only

An ordered array or comma-separated list of wakeword keywords Only necessary when not passing the filter, detect, and encode paths.

Defined in

types.ts:404

Keyword Config

detect

• detect: string | number

The "detect" Tensorflow-Lite model. If specified, filter and encode are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. detect: require('./detect.tflite'))

The encode model is used to perform each autoregressive step over the mel frames; its inputs should be shaped [mel-length, mel-width], and its outputs [encode-width], with an additional state input/output shaped [state-width]

Defined in

types.ts:272

encode

• encode: string | number

The "encode" Tensorflow-Lite model. If specified, filter and detect are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. encode: require('./encode.tflite'))

Its inputs should be shaped [encode-length, encode-width], and its outputs

Defined in

types.ts:283

filter

• filter: string | number

The "filter" Tensorflow-Lite model. If specified, detect and encode are also required.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. filter: require('./filter.tflite'))

The filter model is used to calculate a mel spectrogram frame from the linear STFT; its inputs should be shaped [fft-width], and its outputs [mel-width]

Defined in

types.ts:260

Either metadata or classes is required, and they are mutually exclusive.

metadata

• metadata: string | number

The JSON file for Keyword metadata. Required if keyword.classes is not specified.

This field accepts 2 types of values.

1. A string representing a remote URL from which to download and cache the file (presumably from a CDN).
2. A source object retrieved by a require or import (e.g. metadata: require('./metadata.sjson')).

IMPORTANT: a special extension is used for local metadata JSON files (.sjson) when using require or import so the file is not parsed when included but instead imported as a source object. This makes it so the file is read and parsed by the underlying native libraries instead.

Defined in

types.ts:424

classes

• classes: string | string[]

A comma-separated list or an ordered array of class names for the keywords. The name corresponding to the most likely class will be returned in the transcript field when the recognition event is raised. Required if keyword.metadata is not specified.

Defined in

types.ts:434

Advanced Wakeword and Keyword Config

These properties can be passed to either the wakeword or keyword config object, but are not shared.

encodeLength

• Optional encodeLength: number

advanced

The length of the sliding window of encoder output used as an input to the classifier, in milliseconds

Defined in

types.ts:293

encodeWidth

• Optional encodeWidth: number

advanced

The size of the encoder output, in vector units

Defined in

types.ts:299

fftHopLength

• Optional fftHopLength: number

advanced

The length of time to skip each time the overlapping STFT is calculated, in milliseconds

Defined in

types.ts:322

fftWindowSize

• Optional fftWindowSize: number

advanced

The size of the signal window used to calculate the STFT, in number of samples - should be a power of 2 for maximum efficiency

Defined in

types.ts:306

fftWindowType

• Optional fftWindowType: string

advanced

Android-only

The name of the windowing function to apply to each audio frame before calculating the STFT; currently the "hann" window is supported

Defined in

types.ts:315

melFrameLength

• Optional melFrameLength: number

advanced

The length of time to skip each time the overlapping STFT is calculated, in milliseconds

Defined in

types.ts:329

melFrameWidth

• Optional melFrameWidth: number

advanced

The size of each mel spectrogram frame, in number of filterbank components

Defined in

types.ts:336

preEmphasis

• Optional preEmphasis: number

advanced

The pre-emphasis filter weight to apply to the normalized audio signal (0 for no pre-emphasis)

Defined in

types.ts:343

stateWidth

• Optional stateWidth: number

advanced

The size of the encoder state, in vector units (defaults to wake-encode-width)

Defined in

types.ts:349

threshold

• Optional threshold: number

advanced

The threshold of the classifier's posterior output, above which the trigger activates the pipeline, in the range [0, 1]

Defined in

types.ts:356

---

License

Apache-2.0

Copyright 2021 Spokestack