search

MicrosoftDocs / azure-docs

Open source documentation of Microsoft Azure
https://docs.microsoft.com/azure
Creative Commons Attribution 4.0 International
10.24k stars 21.41k forks source link

Word-level Emphasis only currently works for en-US-DavisNeural #109997

Closed vhhughes closed 1 year ago

vhhughes commented 1 year ago

In the Adjust Emphasis section, it incorrectly states that word-level emphasis tuning works for en-US-GuyNeural and en-US-JaneNeural. It does not. Of the three listed, it only currently works for en-US-DavisNeural. (Looking forward to it working for many more!)

Document Details

⚠ Do not edit this section. It is required for learn.microsoft.com ➟ GitHub issue linking.

AjayBathini-MSFT commented 1 year ago

@vhhughes Thanks for your feedback! We will investigate and update as appropriate.

Naveenommi-MSFT commented 1 year ago

@vhhughes Thank you for bringing this to our attention. I've delegated this to content author @eric-urban, who will review it and offer their insightful opinions.

eric-urban commented 1 year ago

@Naveenommi-MSFT - The feature team tried all 3 voices successfully. Could you please share more details about how to reproduce the issue, such as your Speech resource region and SSML sample? Please also see more Azure Cognitive Services support and help options here. Thanks

vhhughes commented 1 year ago

I'm using a S0 resource in East US. And at the moment, I'm no longer able to see it working with any of the three voices. Below is the SSML I am using. I have tried it with the "strong" level, as well.

`

This is an example of how vocal emphasis can be used to bring the listener's attention to a particular word in a sentence. This is an example of how vocal emphasis can be used to bring the listener's attention to a particular word in a sentence. This is an example of how vocal emphasis can be used to bring the listener's attention to a particular word in a sentence.`
eric-urban commented 1 year ago

@vhhughes - I've heard from our engineering team that emphasis might not be added for some words depending in part on the voice. So, most of the time it's going to work well, but it's not something that is guaranteed. We're going to clarify that in the documentation.

vhhughes commented 1 year ago

Thanks for the follow-up here, Eric.

eric-urban commented 1 year ago

@vhhughes - Here are more details that we added to documentation: "For words that have low pitch and short duration, the pitch might not be raised enough to be noticed." Thanks again!

eric-urban commented 1 year ago

please-close