Issue with Continuous Speech Recognition Returning Omitted Words in Azure Speech Service for recognize words in pt-BR

[gabrielangelo]

Hey, I’m using the Azure Speech Service for continuous speech recognition, following the official JavaScript sample from the cognitive-services-speech-sdk repository. I’ve encountered a behavior I’d like to clarify.

When using the continuous recognition mode, the SDK takes a long time to process, often returning a result with all words omitted. This is not the expected behavior, and I’m seeking clarification on why this happens and if there's a way to address it.

To Reproduce

Follow the official JavaScript sample from the cognitive-services-speech-sdk repository for continuous speech recognition.
Run the continuous recognition with an audio input.
Observe the behavior where the SDK takes an extended amount of time to process and returns a result with all words omitted.

Expected behavior
The expected behavior is that the continuous recognition mode should return accurate transcription of the spoken words without omissions.

Version of the Cognitive Services Speech SDK
[email protected]

Platform, Operating System, and Programming Language
OS: Ubuntu 22.04 LTS
Language: Javascript

Additional context
My config in settings.js

export const serviceRegion =     "brazilsouth"; // e.g., "westus"
export const language =          "pt-BR";

My sample audio details.

The shell