Azure AI Speech
An Azure service that integrates speech processing into apps and services.
How do I change Multilingual Jenny's language in python?
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(38.4, 28.000000000000004%, 13%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EJG%3C/text%3E%3C/svg%3E)
John Garrison
5
Reputation points
I've ran into the 53,435th issue where the documentation doesn't help me. This is becoming frustratingly common. I want to use Multilingual Jenny to speak various languages. I'm using Python to generate the audio. No matter what I set the language to, she speaks in English. I've tried changing just the language and changing her language. Neither of these work. How do I make her speak a different language? According to the documentation, the first one should work. Changing the synthesis language is supposed to change the output language. But it doesn't. However, there's no specific documentation on multilingual Jenny, because of course there isn't.
language = "ko-KR"
voice = "en-US-JennyMultilingualNeural"
speech_config.speech_synthesis_voice_name = voice # Set the desired voice name
speech_config.speech_synthesis_language = language # Set the desired language code
language = "ko-KR"
voice = "ko-KR-JennyMultilingualNeural"
speech_config.speech_synthesis_voice_name = voice # Set the desired voice name
speech_config.speech_synthesis_language = language # Set the desired language code
Azure AI Speech
{count} votes
-
romungi-MSFT 49,096 Reputation points Microsoft Employee Moderator2023-05-26T10:51:47.24+00:00 @John Garrison I understand that it is not obvious from the QuickStart documentation that the support for multilingual voices is only available through SSML. The language support page mentions this as a footnote.
So, using SSML along with speak_ssml_async() will work for this scenario. Here is a sample SSML to use the multilingual voice.
<speak version="1.0" xmlns="https://www.w3.org/2001/10/synthesis" xmlns:mstts="http://www.w3.org/2001/mstts" xmlns:emo="http://www.w3.org/2009/10/emotionml" xml:lang="en-US"> <voice name="Microsoft Server Speech Text to Speech Voice (en-US, JennyMultilingualNeural)"><lang xml:lang="ko-KR"> 안녕 </lang> </voice> <voice name="Microsoft Server Speech Text to Speech Voice (en-US, JennyMultilingualNeural)"><lang xml:lang="es-MX"> Estoy deseando que lleguen las cosas emocionantes </lang> </voice> <voice name="Microsoft Server Speech Text to Speech Voice (en-US, JennyMultilingualNeural)"><lang xml:lang="zh-CN"> 你好 </lang> </voice> </speak>A sample to use this SSML:
speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config) ssml_string = open("ssml.xml", "r").read() # The synthesized speech is expected to be heard on the speaker with this line executed. result = speech_synthesizer.speak_ssml_async(ssml_string).get()For the scenario that you tried, setting of voice and language with speechConfig will work for scenarios where you want to match your input text and use voices which are not multilingual neural i.e
en-US-JennyMultilingualNeural.So, this setting will work based on the following priority.The voice that speaks is determined in order of priority as follows:
- If you don't set
SpeechSynthesisVoiceNameorSpeechSynthesisLanguage, the default voice foren-USwill speak. - If you only set
SpeechSynthesisLanguage, the default voice for the specified locale will speak. - If both
SpeechSynthesisVoiceNameandSpeechSynthesisLanguageare set, theSpeechSynthesisLanguagesetting is ignored. The voice that you specified viaSpeechSynthesisVoiceNamewill speak. - If the voice element is set via Speech Synthesis Markup Language (SSML), the
SpeechSynthesisVoiceNameandSpeechSynthesisLanguagesettings are ignored.
I hope this helps!!
- If you don't set
-
John Garrison 5 Reputation points
2023-05-26T11:25:08.9966667+00:00 Ok. Just to make sure I understand the billing correctly, the charge from you example would only be for:
안녕
Estoy deseando que lleguen las cosas emocionantes
你好
With the third one counting double because it's Chinese characters and the others counted once per character, correct? The documentation mentions some things about controlling the speech also counting as billable. But the language codes don't appear to be on that list. I just want to make sure I understand what I'm getting myself into before I run a script processing a bunch of sentences.
EDIT: That appears to be incorrect. The audio creation tool from the speech portal shows the number of billable characters at 145 when using SSML and 55 when not. Do I really have to pay three times as much to use the multilingual voices?
EDIT2: Playing with a character counter, it appears as though the entirety of the <lang> tags is charged for. Is there a way to set that once and loop through a bunch of sentences with the same language setting using Python, or must it be set every single time I want her to say something in a language other than en_US?
-
romungi-MSFT 49,096 Reputation points Microsoft Employee Moderator
2023-05-26T14:41:05.05+00:00 @John Garrison SSML provides the capability to use other features of voice like rate, pitch, phonemes etc. This makes it expensive compared to just passing the text to voice using speak_text_aysnc(), If your application does not need these capabilities, you can just set the language and voice with speech config and call speak_text_async()
If you have only one language to speak, setting lang once in the SSML should work. It is preferrable to use multilingual voice like Jenny for customers whose use case is to use single voice characteristics across different languages in their applications.
SSML document structure is based on Speech Synthesis Markup Language Version 1.0 so the characters in every request are counted when it is synthesized by the API.
Sign in to comment
Sign in to answer