Version

master (targeting 4.12.0)

Describe the bug

After we bumped to Speech SDK 1.15.0, there are some bugs that we need to add workaround code.

This tracking bug is for the bug that Speech SDK send us wrong audio output format. When we detect it is MP3 using Direct Line Speech, we hardcoded it to RIFF 16kHz mono.

Steps to reproduce

1. Remove the TODO code in playCognitiveServicesStream.js by searching this bug number
2. Talk to the bot through speech

Expected behavior

The bot response should be synthesized at correct sampling rate (16 kHz).

Instead, it synthesized at 24 kHz, which results in a kid-like voice.

Additional context

The workaround should be removed after either 313, or both 313 and 314 is resolved.

microsoft/cognitive-services-speech-sdk-js#313
microsoft/cognitive-services-speech-sdk-js#314 - resolved

[Bug]