Skip to content

Releases: csdcorp/speech_to_text

Android stop fix

23 Aug 16:44
Compare
Choose a tag to compare

New

  • new parameter localeId on the SpeechToTextProvider listen method supports selecting a
    non default languge for the recognizer when using the provider.

Fix

  • A work around for a bug in Android 10(29) that made stop and cancel not work. The plugin now
    calls destroy on the Android SpeechRecognizer which terminates the listen session immediately.
    There is a downside to this approach in that it does not always report results that were in process
    when destroy was called, however it does mean that stop works.
  • The example for SpeechToTextProvider, provider_example.dart had an error that made it always
    use the default language and ignore the drop-down selection. This has been corrected.

Doc update

04 Aug 15:50
Compare
Choose a tag to compare

Fix

  • Updated the version number in the readme file, which I forget to do, every single time.

iOS final Results fix

04 Aug 15:39
Compare
Choose a tag to compare

New

  • new parameter sampleRate on the listen method supports some older iOS devices by allowing
    customization to support the expected hardware sample rate. 44100 works with some older devices.

Fix

  • pauseFor now times out closer to the expected time and more reliably. Duplicate partial results were
    causing it to extend previously.
  • finalResult is now reliably true on the last result on iOS when pauseFor or listenFor timeout or
    stop is called directly. Previously it would only return true when the stop happened almost immediately

Provider

30 Jun 17:23
Compare
Choose a tag to compare

New

  • new parameter onDevice on the listen method enforces on device recognition for sensitive content
  • onSoundLevelChange now supported on iOS
  • added compile troubleshooting help to README.md
  • SpeechToTextProvider is an alternate and simpler way to interact with the SpeechToText plugin.
  • new provider_example.dart example for usage of SpeechToTextProvider.

Fix

  • on iOS handles some conflicts with other applications better to keep speech working after calls for example

iOS sound handling improvements

18 May 13:58
Compare
Choose a tag to compare

New

  • improved error handling and logging in the iOS implementation
  • added general guides for iOS to the README
  • moved stress testing out of the main example
  • iOS now defaults to using the speaker rather than the receiver for start /stop sounds when no headphones

Fix

  • iOS now properly deactivates the audio session when no longer listening
  • start and stop sounds on iOS should be more reliable when available

Error handling on Android

07 Apr 13:50
Compare
Choose a tag to compare

Breaking

  • listenFor now calls stop rather than cancel as this seems like more useful behaviour

Fix

  • Android no longer stops or cancels the speech recognizer if it has already been shutdown by a
    timeout or other platform behaviour.
  • Android no longer tries to restart the listener when it is already active
  • Now properly notifies errors that happen after listening stops due to platform callback rather than
    client request. See #51

Separate timeout errors

11 Mar 12:57
Compare
Choose a tag to compare
  • error_timeout has been separated into error_network_timeout and error_speech_timeout

Partial results & cancel on error

11 Mar 12:56
Compare
Choose a tag to compare
  • hasPermission to check for the current permission without bringing up the system dialog
  • listen has a new optional cancelOnError parameter to support automatically canceling
    a listening session on a permanent error.
  • listen has a new optional partialResults parameter that controls whether the callback
    receives partial or only final results.

Multiple transcriptions

26 Jan 20:46
Compare
Choose a tag to compare

New

  • speech recognizer now exposes multiple possible transcriptions for each recognized speech
  • alternates list on SpeechRecognitionResult exposes alternate transcriptions of voice
  • confidence on SpeechRecognitionResult gives an estimate of confidence in the transcription
  • isConfident on SpeechRecognitionResult supports testing confidence
  • hasConfidenceRating on SpeechRecognitionResult indicates if confidence was provided from the device
  • new SpeechRecognitionWords class gives details on per transcription words and confidence

Fix

  • speechRecognizer availabilityDidChange was crashing if invoked due to an invalid parameter type
  • Added iOS platform 10 to example Podfile to resolve compilation warnings

Flutter 1.12.0 - Swift update

03 Jan 20:42
Compare
Choose a tag to compare

0.7.2

Breaking

  • Upgrade Swift to version 5 to match Flutter. Projects using this plugin must now switch to 5.