Search Issue Tracker

Won't Fix

Votes

0

Found in

2018.4

2019.3.4f1

2020.2

Issue ID

1227088

Regression

No

[Windows] Timeout is not triggered when DictationRecognizer stops recognizing words

Windows

-

Reproduction steps:
1. Open the attached project ("DictationError.zip")
2. Open the repro scene ("SampleScene")
3. Point your microphone at a speaker that's playing audio of someone speaking, like a podcast
4. Enter Play Mode
5. Let the dication go for a few minutes
6. Inspect the Console Logs afterward

Expected result: Longest time since last recognition is under 2-3 seconds
Actual result: Longest time since last recognition goes up to 30s and more; in those intervals, speech is happening, but not being dictated

Reproducible with: 2018.4.25f1, 2019.4.4f1, 2020.1.0b16, 2020.2.0a18

  1. Resolution Note (2020.2.X):

    The 30 second delay is coming from the Windows SpeechContinuousRecognition API (https://docs.microsoft.com/en-us/uwp/api/Windows.Media.SpeechRecognition.SpeechContinuousRecognitionSession?view=winrt-19041) and therefore is a Microsoft bug which we can't fix.

    Unlike KeywordRecognizer, this API uses a web service to process Speech-to-Text in the "Cloud", and it appears the Microsoft server sometimes doesn't respond appropriately. The best work around is to implement your own timeout: if there's no response during a dictation session after ~10 seconds, then call Stop() on the session and then Start() to being a new session.

Add comment

Log in to post comment

All about bugs

View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.