speech_to_text

pub package build status codecov

A library that exposes device specific speech recognition capability.

This plugin contains a set of classes that make it easy to use the speech recognition capabilities of the underlying platform in Flutter. It supports Android, iOS and web. The target use cases for this library are commands and short phrases, not continuous spoken conversion or always on listening.

Recent Updates

The 4.2.0 version is significantly faster starting to listen on iOS (~500 ms) and makes null safety the default release.

The 4.0.0 version adds null safety support thanks to the fabulous work of @kaladron. Support for web is also included though it is not yet well tested.

The 3.0.0 version uses the newer style of plugin development based on a common platform interface package. This will make it easier to support web and other platforms.

Note: Feedback from any test devices is welcome.

Using

To recognize text from the microphone import the package and call the plugin, like so:

import 'package:speech_to_text/speech_to_text.dart' as stt;

    stt.SpeechToText speech = stt.SpeechToText();
    bool available = await speech.initialize( onStatus: statusListener, onError: errorListener );
    if ( available ) {
        speech.listen( onResult: resultListener );
    }
    else {
        print("The user has denied the use of speech recognition.");
    }
    // some time later...
    speech.stop()

Initialize once

The initialize method only needs to be called once per application session. After that listen, start, stop, and cancel can be used to interact with the plugin. Subsequent calls to initialize are ignored which is safe but does mean that the onStatus and onError callbacks cannot be reset after the first call to initialize. For that reason there should be only one instance of the plugin per application. The SpeechToTextProvider is one way to create a single instance and easily reuse it in multiple widgets.

Permissions

Applications using this plugin require user permissions.

iOS

Add the following keys to your Info.plist file, located in <project root>/ios/Runner/Info.plist:

  • NSSpeechRecognitionUsageDescription - describe why your app uses speech recognition. This is called Privacy - Speech Recognition Usage Description in the visual editor.
  • NSMicrophoneUsageDescription - describe why your app needs access to the microphone. This is called Privacy - Microphone Usage Description in the visual editor.

Android

Add the record audio permission to your AndroidManifest.xml file, located in <project root>/android/app/src/main/AndroidManifest.xml.

  • android.permission.RECORD_AUDIO - this permission is required for microphone access.
  • android.permission.INTERNET - this permission is required because speech recognition may use remote services.

Android SDK 30 or later

If you are targeting Android SDK, i.e. you set your targetSDKVersion to 30 or later, then you will need to add the following to your AndroidManifest.xml right after the permissions section. See the example app for the complete usage.

<queries>
    <intent>
        <action android:name="android.speech.RecognitionService" />
    </intent>
</queries>

Adding Sounds for iOS (optional)

Android automatically plays system sounds when speech listening starts or stops but iOS does not. This plugin supports playing sounds to indicate listening status on iOS if sound files are available as assets in the application. To enable sounds in an application using this plugin add the sound files to the project and reference them in the assets section of the application pubspec.yaml. The location and filenames of the sound files must exactly match what is shown below or they will not be found. The example application for the plugin shows the usage. Note These files should be very short as they delay the start / end of the speech recognizer until the sound playback is complete.

  assets:
  - assets/sounds/speech_to_text_listening.m4r
  - assets/sounds/speech_to_text_cancel.m4r
  - assets/sounds/speech_to_text_stop.m4r
  • speech_to_text_listening.m4r - played when the listen method is called.
  • speech_to_text_cancel.m4r - played when the cancel method is called.
  • speech_to_text_stop.m4r - played when the stop method is called.

Tips

Switching Recognition Language

The speech_to_text plugin uses the default locale for the device for speech recognition by default. However it also supports using any language installed on the device. To find the available languages and select a particular language use these properties.

There's a locales property on the SpeechToText instance that provides the list of locales installed on the device as LocaleName instances. Then the listen method takes an optional localeId named param which would be the localeId property of any of the values returned in locales. A call looks like this:

   var locales = await speech.locales();

   // Some UI or other code to select a locale from the list
   // resulting in an index, selectedLocale

   var selectedLocale = locales[selectedLocale];
   speech.listen(
       onResult: resultListener,
       localeId: selectedLocale.localeId,
       );

Troubleshooting

Speech recognition stops after a brief pause on Android

Android speech recognition has a very short timeout when the speaker pauses. The duration seems to vary by device and version of the Android OS. In the devices I've used none have had a pause longer than 5 seconds. Unfortunately there appears to be no way to change that behaviour.

Continuous speech recognition

There have been a number of questions about how to achieve continuous speech recognition using this plugin. Currently the plugin is designed for short intermittent use, like when expecting a response to a question, or issuing a single voice command. Issue #63 is the current home for that discussion. There is not yet a way to achieve this goal using the Android or iOS speech recognition capabilities.

There are at least two separate use cases for continuous speech recognition:

  1. voice assistant style, where recognition of a particular phrase triggers an interaction;
  2. dictation of text for input.

Voice assistant style interaction is possibly better handled by integrating with the existing assistant capability on the device rather than building out a separate capability. Text dictation is available through the keyboard for standard text input controls though there are other uses of dictation that are not currently well supported.

Browser support for speech recognition

Web browsers vary in their level of support for speech recognition. This issue has some details. The best lists I've seen are https://caniuse.com/speech-recognition and https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognition. In particular in issue #239 it was reported that Brave Browser and Firefox for Linux do not support speech recognition.

Speech recognition from recorded audio

There have been a number of questions about whether speech can be recognized from recorded audio. The short answer is that this may be possible on iOS but doesn't appear to be on Android. There is an open issue on this here #205.

SDK version error trying to compile for Android

Manifest merger failed : uses-sdk:minSdkVersion 16 cannot be smaller than version 21 declared in library [:speech_to_text]

The speech_to_text plugin requires at least Android SDK 21 because some of the speech functions in Android were only introduced in that version. To fix this error you need to change the build.gradle entry to reflect this version. Here's what the relevant part of that file looked like as of this writing:

   defaultConfig {
        applicationId "com.example.app"
        minSdkVersion 21
        targetSdkVersion 28
        versionCode flutterVersionCode.toInteger()
        versionName flutterVersionName
        testInstrumentationRunner "androidx.test.runner.AndroidJUnitRunner"
    }

Recording audio on Android

It is not currently possible to record audio on Android while doing speech recognition. The only solution right now is to stop recording while the speech recognizer is active and then start again after.

Incorrect Swift version trying to compile for iOS

/Users/markvandergon/flutter/.pub-cache/hosted/pub.dartlang.org/speech_to_text-1.1.0/ios/Classes/SwiftSpeechToTextPlugin.swift:224:44: error: value of type 'SwiftSpeechToTextPlugin' has no member 'AVAudioSession'
                rememberedAudioCategory = self.AVAudioSession.Category
                                          ~~~~ ^~~~~~~~~~~~~~
    /Users/markvandergon/flutter/.pub-cache/hosted/pub.dartlang.org/speech_to_text-1.1.0/ios/Classes/SwiftSpeechToTextPlugin.swift:227:63: error: type 'Int' has no member 'notifyOthersOnDeactivation'
                try self.audioSession.setActive(true, withFlags: .notifyOthersOnDeactivation)

This happens when the Swift language version is not set correctly. See this thread for help https://github.com/csdcorp/speech_to_text/issues/45.

Swift not supported trying to compile for iOS

`speech_to_text` does not specify a Swift version and none of the targets (`Runner`) integrating it have the `SWIFT_VERSION` attribute set.

This usually happens for older projects that only support Objective-C. See this thread for help https://github.com/csdcorp/speech_to_text/issues/88.

Not working on a particular Android device

The symptom for this issue is that the initialize method will always fail. If you turn on debug logging using the debugLogging: true flag on the initialize method you'll see 'Speech recognition unavailable' in the Android log. There's a lengthy issue discussion here https://github.com/csdcorp/speech_to_text/issues/36 about this. The issue seems to be that the recognizer is not always automatically enabled on the device. Two key things helped resolve the issue in this case at least.

Not working on an Android emulator

The above tip about getting it working on an Android device is also useful for emulators. Some users have reported seeing another error on Android simulators - sdk gphone x86 (Pixel 3a API 30). AUDIO_RECORD perms were in Manifest, also manually set Mic perms in Android Settings. When running sample app, Initialize works, but Start failed the log looks as follows.

D/SpeechToTextPlugin(12555): put partial
D/SpeechToTextPlugin(12555): put languageTag
D/SpeechToTextPlugin(12555): Error 9 after start at 35 1000.0 / -100.0
D/SpeechToTextPlugin(12555): Cancel listening

Resolved by

Resolved it by Opening Google, clicking Mic icon and granting it perms, then everything on the App works...

First

  1. Go to Google Play
  2. Search for 'Google'
  3. You should find this app: https://play.google.com/store/apps/details?id=com.google.android.googlequicksearchbox If 'Disabled' enable it

This is the SO post that helped: https://stackoverflow.com/questions/28769320/how-to-check-wether-speech-recognition-is-available-or-not

Second

Ensure the app has the required permissions. The symptom for this that you get a permanent error notification 'error_audio_error` when starting a listen session. Here's a Stack Overflow post that addresses that https://stackoverflow.com/questions/46376193/android-speechrecognizer-audio-recording-error Here's the important excerpt:

You should go to system setting, Apps, Google app, then enable its permission of microphone.

iOS recognition guidelines

Apple has quite a good guide on the user experience for using speech, the original is here https://developer.apple.com/documentation/speech/sfspeechrecognizer This is the section that I think is particularly relevant:

Create a Great User Experience for Speech Recognition

Here are some tips to consider when adding speech recognition support to your app.

Be prepared to handle failures caused by speech recognition limits. Because speech recognition is a network-based service, limits are enforced so that the service can remain freely available to all apps. Individual devices may be limited in the number of recognitions that can be performed per day, and each app may be throttled globally based on the number of requests it makes per day. If a recognition request fails quickly (within a second or two of starting), check to see if the recognition service became unavailable. If it is, you may want to ask users to try again later.

Plan for a one-minute limit on audio duration. Speech recognition places a relatively high burden on battery life and network usage. To minimize this burden, the framework stops speech recognition tasks that last longer than one minute. This limit is similar to the one for keyboard-related dictation. Remind the user when your app is recording. For example, display a visual indicator and play sounds at the beginning and end of speech recognition to help users understand that they're being actively recorded. You can also display speech as it is being recognized so that users understand what your app is doing and see any mistakes made during the recognition process.

Do not perform speech recognition on private or sensitive information. Some speech is not appropriate for recognition. Don't send passwords, health or financial data, and other sensitive speech for recognition.

Libraries

speech_recognition_error
speech_recognition_event
speech_recognition_result
speech_to_text
speech_to_text_provider
speech_to_text_web