AzureTTS
AzureTTS module is a multilingual TTS module of Greta platform, powered by Microsoft Azure Speech Service.
Please refer the supported languages from here.
There are still several technical limitations, please check known issue section of this page.
-
Make sure you have installed python3 and conda. Usually you can install both at once by installing anaconda.
-
You need to create Azure account and create a Speech resource. Please refer prerequisite of this webpage
-
Create 'SPEECH_KEY.txt' and 'SPEECH_REGION.txt' in bin/Common/Data/AzureTTS and put your credential information into the corresponding text files.
-
Place AzureTTS module from the menu bar of the Modular.jar (Add -> TTS -> AzureTTS) and connect to TTS module as shown in the screenshot bellow.
-
From the AzureTTS window, select your favorite voice (You can add other choices by the additional steps explained later in this page).
-
If you want to use language which is not competible with UTF-8 encoding (e.g. French), you need to select Greta's character language as well. You can change it from CharacterManager window. At this moment, we added exception handling only for French (ISO-8859-1). If you want to add other exception handlings, please check and modify this file.
-
Now, you can send any message as same as other TTS modules, but with your favorite language!
- Voice generation time is longer than other TTS (minimum about 1 second)
- Small pause might be inserted at the point of each TimeMarker (especially troublesome for word-level TimeMarkers)
- Some voices don't support mouth shape (viseme) generation, which resulted in no mouth movement (please check here)
- Frame implementation file
- Main implementation file
- Actual call is based on python programs here
