This article is contributed. See the original author and article here.

This post was co-authored with Qinying Liao, Sheng Zhao, Gang Wang, Yueying Liu


 


Neural Text to Speech (Neural TTS), a powerful speech synthesis capability of Cognitive Services on Azure, enables you to convert text to lifelike speech which is close to human-parity.  Since its launch, we have seen it widely adopted in a variety of scenarios by many Azure customers, from voice assistants to audio content creation. More and more customers are asking for richer and more diverse choices of synthetic voices for different use cases.


 


Today, we are excited to announce that Azure Neural TTS has added 51 new voices for a total of 129 neural voices across 54 languages/locales. With this release, we provide at least one male and one female voice for customers to choose in each language/locale.  In total, Azure TTS now enables developers to reach millions more people with more than 200 voices available in standard and neural TTS.


 


What’s new


 


Neural TTS has now been extended to support 51 new voices, which will bring to you the capability to have both male and female voices in each language for your apps. You can hear samples of the voices below, or try them with your own text in our demo.


 


46 new voices are generally available


 


In total 46 new voices are released across the 49 locales that are generally available in the Azure data centers/regions that support neural TTS (see the full list of Azure regions here).


 















































































































































































































































































































































Locale



Language 



Gender



Voice



Sample audio



ar-EG



Arabic (Egypt)



Male



ShakirNeural



البركان هو أكثر ما في الطبيعــة إثارة للرهبة



ar-SA



Arabic (Saudi Arabia)



Male



HamedNeural



الناس مَعادن، تصدأ بالملل، وتتمدد بالأمل، وتنكمش بالألم



bg-BG



Bulgarian (Bulgary)



Male



BorislavNeural



Шофьорът задължително трябва да вземе експерт за второ мнение, за да провери дали всички системи на автомобила работят нормално.



ca-ES



Catalan (Spain)



Male



EnricNeural



Les activitats docents tenen lloc al campus del Poblenou.



ca-ES



Catalan (Spain)



Female



JoanaNeural



L’artista està considerat com el pintor de les multituds.



cs-CZ



Czech (Czech)



Male



AntoninNeural



Opravdový zasvěcenec ví, že nejmocnějším tajemstvím je to, které nemá žádný obsah.



da-DK



Danish (Denmark)



Male



JeppeNeural



61 procent af de kandidatstuderende er kvinder.



de-AT



German (Austria)



Male



JonasNeural



Das ist das letzte lange Pfingstwochenende für Schülerinnen und Schüler.



de-CH



German (Switzerland)



Male



JanNeural



Eine Person, die sich bei Brandausbruch im oberen Stock aufgehalten hat, hat sich noch rechtzeitig in Sicherheit bringen können.



el-GR



Greek (Greece)



Male



NestorasNeural



Συγκλονιστικές εξελίξεις και ανατροπές στα επόμενα επεισόδια .



en-CA



English (Canada)



Male



LiamNeural



He had held the position since 2010.



en-IE



English (Ireland)



Male



ConnorNeural



Life is short, think before you talk.



en-IN



English (India)



Male



PrabhatNeural



Sometimes you can see snow on the mountains.



fi-FI



Finnish (Finland)



Male



HarriNeural



Yhtiö kertoi loppuvuoden tuloksestaan ennakkotietoja.



fi-FI



Finnish (Finland)



Female



SelmaNeural



Hevoset ovat uljaita ja nopeita eläimiä.



fr-CH



French (Switzerland)



Male



FabriceNeural



La Suisse comptera 5,6 millions (12%) de personnes actives en 2050.



he-IL



Hebrew (Israel)



Male



AvriNeural



הוא אמר שהמספרים מדאיגים בשל עצמם, אבל בכל הישיבות שלנו המסקנה היא שזה סימפטום למשהו רחב יותר.



hi-IN



Hindi (India)



Male



MadhurNeural



सिद्धार्थ ने भी शहनाज के साथ इस इवेंट की फोटो शेयर की है।



hr-HR



Croatian (Croatia)



Male



SreckoNeural



Video je pregledan gotovo 70 tisuća puta, a neki od obožavatelja su mu u komentarima pisali kako ih je motivirao.



hu-HU



Hungarian (Hungary)



Male



TamasNeural



A lakóhelyem nagyon komfortos.



id-ID



Indonesian (Indonesia)



Female



GadisNeural



Inflasi ringan terjadi apabila kenaikan harga berada di bawah angka 10% setahun.



ms-MY



Malay (Malaysia)



Male



OsmanNeural



Setiap individu perlu memakai topeng muka ketika berada di luar.



nb-NO



Norwegian (Bokmål, Norway)



Male



FinnNeural



Jansson forteller at den svenske øya tar imot rundt 8000 besøkende fra Norge årlig.



nb-NO



Norwegian (Bokmål, Norway)



Female



PernilleNeural



For en fantastisk forestilling!



nl-NL



Dutch (Netherlands)



Female



FennaNeural



De afstand tussen Rotterdam en Breda is ongeveer 45 km.



nl-NL



Dutch (Netherlands)



Male



MaartenNeural



Zij heeft haar studie al een tijdje geleden afgerond.



pl-PL



Polish (Poland)



Female



AgnieszkaNeural



To już nie będzie to samo, będzie drożej.



pl-PL



Polish (Poland)



Male



MarekNeural



Na wszelki wypadek sprawdź, czy coś cię jednak nie zaskoczy.



pt-PT



Portuguese (Portugal)



Male



DuarteNeural



Para a aprovação do exame, tenho de ter pelo menos 80% das respostas corretas.



pt-PT



Portuguese (Portugal)



Female



RaquelNeural



A minha mãe ensinou-me que devo ter respeito por todos, mas principalmente pelos mais velhos.



ro-RO



Romanian (Romania)



Male



EmilNeural



Actul normativ se axează pe instituirea de măsuri active, 41,5 % din salariul de bază la revenirea din șomaj tehnic.



ru-RU



Russian (Russia)



Male



DmitryNeural



Ранее посольство требовало от агентства опровержения статьи о количестве больничных коек в России.



ru-RU



Russian (Russia)



Female



SvetlanaNeural



Изменений в организме людей, попробовавших еду без приправ, не произошло.



sk-SK



Slovak (Slovakia)



Male



LukasNeural



Zápis 45 % je v skutočnosti iba skratka pre zlomok.



sl-SI



Slovenian (Slovenia)



Male



RokNeural



Zloraba bonov in dvigovanje cen turističnih storitev je nesprejemljivo ravnanje.



sv-SE



Swedish (Sweden)



Male



MattiasNeural



Båda lagen bjöd på riktigt bra hockey och skapade flera riktigt bra målchanser.



sv-SE



Swedish (Sweden)



Female



SofieNeural



Det fanns ingen trafik runt torget.



ta-IN



Tamil (India)



Male



ValluvarNeural



எவ்வளவு அருமையான பாடல் அது!



te-IN



Telugu (India)



Male



MohanNeural



అబ్బ, ఎంత పెద్ద భవనమో!



th-TH



Thai (Thailand)



Male



NiwatNeural



ธุรกิจขายอาหารเป็นธุรกิจที่ได้รับความนิยมมากที่สุด



tr-TR



Turkish (Turkey)



Male



AhmetNeural



Sosyal mesafeye büyük ölçüde riayet eden çocuklar, başta mahalle parkları olmak üzere sahiller ve oyun parklarında enerji attı.



vi-VN



Vietnamese (Vietnam)



Male



NamMinhNeural



Nhiệt độ hiện tại ở thành phố Hồ Chí Minh là 38 độ C.



zh-HK



Chinese (Cantonese, Traditional)



Female



HiuMaanNeural



抗疫舉措成為安全重啟經濟的重要一環。



zh-HK



Chinese (Cantonese, Traditional)



Male



WanLungNeural



隨着疫情緩和,愈來愈多人回到辦公室上班,但是很多人仍想留在家中工作(work from home)。



zh-TW



Chinese (Taiwanese Mandarin)



Female



HsiaoChenNeural



賭博的勝率應該不到50%。



zh-TW



Chinese (Taiwanese Mandarin)



Male



YunJheNeural



台北車站大廳能不能坐,連日引發正反意見。



 


 


5 new voices are in public preview


 


We have also added 5 male voices in the 5 low-resource languages that have been supported since November. These voices are available in public preview in three Azure regions: EastUS, SouthEastAsia and WestEurope.


 


Hear the samples below:
















































Locale



Language



Gender



Voice Name



Sample audio



et-EE



Estonian (Estonia)



Male



KertNeural



Ametlikku meetodit sellise pettuse avastamiseks ei olegi olemas.



ga-IE



Irish (Ireland)



Male



ColmNeural



Ritheadh próiseas comhairliúcháin faoin scéal sa bhfómhar.



lt-LT



Lithuanian (Lithuania)



Male



LeonasNeural



Aišku, anksčiau ar vėliau paaiškės tos priežastys.



lv-LV



Latvian (Latvia)



Male



NilsNeural



Aizvadīto gadu uzņēmums noslēdzis ar 6,3 miljonu eiro zaudējumiem.



mt-MT



Maltese (Malta)



Male



JosephNeural



Anki tfajjel tal-primarja jaf li l-popolazzjoni tikber fejn hemm il-prosperità.



 


With this release, we now support a total of 129 neural voices across 54 languages/locales. In addition, over 70 standard voices are available in 49 languages/locales. Visit Language support – Speech service – Azure Cognitive Services | Microsoft Docs for the full language and voice list.


garheMS_5-1608027137739.png


 


Continuous voice quality improvement


 


In general, Neural TTS can convert text to lifelike speech, however there are nuances that can always be improved. For example, we have customers who have requested the ability to support a scenario where Katja, our de-DE neural voice, can pronounce English words in the context of a German sentence. This was valuable feedback, and we anticipate a similar need across languages.


 


For German, we observed that our users prefer the voice to handle an English word/phrase as close as the native English pronunciation. To enable a voice model to speak English as a second language, it is normally required that we collect the speech data of the same speaker speaking English besides his/her native language. This is a big challenge as we do not have sufficient multi-language speech data from our German voice talents. By leveraging cross-lingual capability of UNI-TTS, we are able to generate more English pronunciation data with the transferred voice from our German voice talent. Such data is used to improve the quality of the English word/phrase pronunciations for the German Katja voice, so Katja can pronounce English words in a more natural way.


 


The CMOS metric is used to measure the improvement of the English word pronunciation for Katja. Below table shows that the updated model is significantly better in pronouncing English words in the context of a German sentence.


 




















Script Old New
Star Wars – Das Erwachen der Macht




Three Billboards outside Ebbing, Missouri.





 


This improvement has now been released to the Azure Neural TTS service for Katja. Moving forward, we’ll extend this capability to support more languages.


 


Tell us your experience!


 


By offering more voices across more languages and locales, we anticipate developers across the world will be able to build applications that change experiences for millions. Whether you’re building a voice-enabled chatbot or IoT device, an IVR solution, adding read-aloud features to your app, converting e-books to audio books, or even adding Speech to a translation app, you can make all these experiences natural sounding and fun with Neural TTS.


 


Let us know how you are using or plan to use Neural TTS voices in this form. If you prefer, you can also contact us at mstts [at] microsoft.com. We look forward to hearing your experience and developing more compelling services together with you for the developers around the world.


 


Get started


Explore the available voices in this demo


Add voice to your app in 15 minutes


Build a voice-enabled bot


Deploy Azure TTS voices on prem with Speech Containers


Build your custom voice

Brought to you by Dr. Ware, Microsoft Office 365 Silver Partner, Charleston SC.