Azure Cognitive Services releases new languages and voices for Neural Text-to-Speech
Published Nov 09 2022 12:28 AM 15.1K Views
Microsoft

This post is co-authored with Melinda Ma, Nick Zhao, Qinying Liao, Gang Wang, Binggong Ding and Sheng Zhao

 

Neural Text-to-Speech (Neural TTS), a powerful speech synthesis capability of Azure Cognitive Services, enables developers to convert text to lifelike speech using AI. Businesses utilize Neural TTS for voice assistants, content read aloud capabilities, accessibility tools, and more. The Azure TTS product team is continuously working on bringing new languages to the world.

 

We are glad to announce that two new languages are introduced to the neural TTS portfolio. With that update, we now support 147 languages/variances in total. In addition, 46 new prebuilt voices are available in preview for a growing list of languages, besides a set of new emotions enabled to many existing voices. That adds up to 449 voices in the neural TTS prebuilt voice family. 

 

With other Cognitive Services including Speech-to-Text, OCR and Translator extended to 100+ languages, Azure AI is one big step closer to its ambition to empower every organization and everyone on the planet to achieve more, without any language barriers.

 

GarfieldHe_0-1668570782716.png

 

2 new languages are generally available

While there are thousands of spoken languages in the world, the top 1% (~70 language) accounts for 80%+ of the global population. Within just a few years of development, the TTS technology has been available in many commonly spoken languages and has been updated in these languages with many advanced features available. However, due to the lack of training data or business needs, there is almost nothing available for those less spoken languages. Inspired by the vision to remove the language barrier for everyone and powered by the low resource setting TTS technology, we keep working to expand our capability to support those less touched languages in the world.

 

Today, we are adding 2 new languages to our neural TTS portfolio. This is just a small portion compared to our goal, but we are on the way!

 

Check out how the voices in these languages sound like with samples below:

Locale

Language

Voice name

Gender

Script

Audio

eu-ES-

Basque

AinhoaNeural

Female

Euskaltzaindiak hiri honetarako onartutako euskal izen bakarra Bilbo da.

eu-ES-

Basque

AnderNeural

Male

Neguko Olinpiar Jokoak neguko kirolak aurkezteko sortu ziren.

hy-AM

Armenian (Armenia)

AnahitNeural

Female

Իսկ ի՞նչն եք ամենից շատ հավանել Հայաստանում:

hy-AM

Armenian (Armenia)

HaykNeural

Male

Կանադայի հոկեյի հավաքականը դարձավ աշխարհի չեմպիոն։

 

2 new variances of Chinese are in public preview

We are glad to introduce two new language variances in Chinese that are now in public preview: Chinese (Wu, Simplified) and Chinese (Cantonese, Simplified). Hear how they go with samples below: 

 

Locale Language Voice name Gender Script Audio
wuu-CN Chinese (Wu,
Simplified)
XiaotongNeural Female 好额,全程24公里,预计通行辰光30分钟,请系好安全带。准备出发。
wuu-CN Chinese (Wu, Simplified) YunzheNeural Male 检测到侬有些疲劳,前方1公里有休息区,建议停车休息,注意安全驾驶。
yue-CN Chinese (Cantonese, Simplified) XiaominNeural Female 苏州市今日阴,气温20℃到26℃。泥水会整污糟你部爱车,唔适合洗车。
yue-CN Chinese (Cantonese, Simplified) YunsongNeural Male 建议你喺车辆无法正常断电时使用该功能,请确认是否关闭整车电源?

Discover the full list of supported languages for Neural Text to Speech, in addition to Microsoft Edge Read aloud.

 

46 new voices are in preview for some popular languages

For each language that we support with prebuilt neural TTS voices, we have provided at least one female and one male voice per locale. However, in the real world, there are scenarios that require one or more voices to reflect diversity and natural conversation. To directly address customer feedback, we work rapidly to bring a richer choice of voices in different languages. In this release, we are starting to introduce 46 new voices in preview for English (Australia), Spanish (Spain), Korean (Korea) and Japanese (Japan).

 

These new voices cover different personas and age groups, which can bring more choices and variety for business scenarios. These voices now are in public preview, available in 3 regions: East US, West Europe and Southeast Asia.

 

We encourage you to try the new languages and voices below. Feedback is welcomed to help inform which voices will be made for General Availability in all regions, depending on customer satisfaction.

 

Locale

Language

Voice name

Gender

Script

Audio

en-AU

English (Australia)

AnnetteNeural

Female

They each accounted for less than 5% of the urinary metabolites.

en-AU

English (Australia)

CarlyNeural

Female

What a lovely bouquet of flowers!

en-AU

English (Australia)

DarrenNeural

Male

I have something here for little Edward.

en-AU

English (Australia)

DuncanNeural

Male

I have come to ask your pardon.

en-AU

English (Australia)

ElsieNeural

Female

Click through to learn more!

en-AU

English (Australia)

FreyaNeural

Female

You need to use about 10 grammes of sugar.

en-AU

English (Australia)

JoanneNeural

Female

The programme teaches you how to eventually run 5 km with ease.

en-AU

English (Australia)

KenNeural

Male

The development site is situated within a new 12 kilometre touristic zone.

en-AU

English (Australia)

KimNeural

Female

It's almost as if it is a lifetime goal.

en-AU

English (Australia)

NeilNeural

Male

I think you have been asleep.

en-AU

English (Australia)

TimNeural

Male

You need to use about 10 grammes of sugar.

en-AU

English (Australia)

TinaNeural

Female

It is a beautiful day, but yesterday it was cold.

es-ES

Spanish (Spain)

AbrilNeural

Female

Tu cumpleaños es el once de noviembre.

es-ES

Spanish (Spain)

ArnauNeural

Male

Desde mañana lunes, se podrán celebrar reuniones al aire libre entre seis personas.

es-ES

Spanish (Spain)

DarioNeural

Male

En dos días el 70 % pasará a la fase 2.

es-ES

Spanish (Spain)

EliasNeural

Male

El Ejecutivo tiene un plan de recuperación del turismo.

es-ES

Spanish (Spain)

EstrellaNeural

Female

Los establecimientos tendrán que firmar una declaración responsable de compromiso.

es-ES

Spanish (Spain)

IreneNeural

Female

Su portátil es de marca H P.

es-ES

Spanish (Spain)

LaiaNeural

Female

Eso no significa que no vaya a haber campaña.

es-ES

Spanish (Spain)

LiaNeural

Female

¿Es la digitalización la clave de nuestro futuro?

es-ES

Spanish (Spain)

NilNeural

Male

Además, está planteando emprender acciones legales.

es-ES

Spanish (Spain)

SaulNeural

Male

Ese es uno de sus grandes objetivos.

es-ES

Spanish (Spain)

TeoNeural

Male

Empezó a trabajar en esta empresa el 17 de septiembre de 2015.

es-ES

Spanish (Spain)

TrianaNeural

Female

El denunciante aseguró que el detenido transportaba un envío.

es-ES

Spanish (Spain)

VeraNeural

Female

Los precios han subido un 30 % desde el año pasado.

ja-JP

Japanese (Japan)

AoiNeural

Female

冬の雪山でスノーボードして休暇を楽しみたい。

ja-JP

Japanese (Japan)

DaichiNeural

Male

ジョギングするために、必要な服装を準備する。

ja-JP

Japanese (Japan)

MayuNeural

Female

植物を植えるのは面白い。

ja-JP

Japanese (Japan)

NaokiNeural

Male

冷蔵庫が壊れたのでカスタマサービスに電話する。

ja-JP

Japanese (Japan)

ShioriNeural

Female

流行りのアプリを使ってみよう。

ko-KR

Korean (Korea)

BongJinNeural

Male

동작을 최소화하는 것도 필요하다.

ko-KR

Korean (Korea)

GookMinNeural

Male

내일 저녁 6시에 회의를 진행 할 예정입니다.

ko-KR

Korean (Korea)

JiminNeural

Female

자극의 위치가 크게 바뀌었다는 사실을 발견했다.

ko-KR

Korean (Korea)

SeoHyeonNeural

Female

마라톤 동호인들이 참여하였다.

ko-KR

Korean (Korea)

SoonBokNeural

Female

평균 37%의 완화도를 보인 걸로 나타났다.

ko-KR

Korean (Korea)

YuJinNeural

Female

충분한 영양 섭취는 건강에 중요하다

 

New styles / emotions are available in more voices

Customers often request the voices to be able to express different styles and emotions based on different content. To address this most frequent ask, we are bringing the multi-style capability to more and more voices and languages.

 

Here are some voices that we have recently enabled to perform a more casual chat style or a more cheerful emotional style, besides their default general tone.

Locale

Language

Voice name

Gender

Style

Script

Audio

en-GB

English (UK)

RyanNeural

Male

Chat

I can chill out a little in such a hot summer. Are you good at swimming?

en-GB

English (UK)

RyanNeural

Male

Cheerful

I'd be happy knowing you're safe.

en-GB

English (UK)

SoniaNeural

Female

Cheerful

I am as happy as Larry because my sister is coming to see me.

en-GB

English (UK)

SoniaNeural

Female

Sad

Jane’s heart sank when she found out that she had lost her job.

es-MX

Spanish (Mexico)

JorgeNeural

Male

Chat

Platica conmigo cuando tengas tiempo.

es-MX

Spanish (Mexico)

JorgeNeural

Male

Cheerful

Carla es una chica muy alegre y siempre sonríe.

fr-FR

French (France)

HenriNeural

Male

Cheerful

Elle est radieuse, la grossesse lui va bien.

fr-FR

French (France)

HenriNeural

Male

Sad

Je ne sais pas pourquoi, mais j’ai le cafard.

it-IT

Italian (Italy)

IsabellaNeural

Female

Chat

Okay, mi va bene. In quel periodo ho anche io le vacanze. Avete già prenotato i biglietti?

it-IT

Italian (Italy)

IsabellaNeural

Female

Cheerful

Il colloquio è andato alla grande, sono proprio soddisfatto.

 

New voices and styles for American English and other languages are generally available

In May 2022, we announced 5 new voices and 10 styles on en-US in preview offering, and later in June 2022, we introduced more voices for it-IT, pt-BR and es-MX for preview. After several months' evaluation, these voices and styles are stable in service and had been used by many customers in their user cases. And we are glad to bring these voices and styles generally available to all service regions.

  • en-US voices: DavisNeural, JaneNeural, JasonNeural, NancyNeural, TonyNeural. And the styles supported include Cheerful, Sad, Angry, Hopeful, Friendly, Unfriendly, Terrified, Excited, Whisphering and Shouting.
  • es-MX voices: CecilioNeural, GerardoNeural, LibertoNeural, LucianoNeural, PelayoNeural, YagoNeural, BeatrizNeural, CarlotaNeural, NuriaNeural, CandelaNeural, LarissaNeural, RenataNeural, MarinaNeural
  • it-IT voices: PierinaNeural, FabiolaNeural, ImeldaNeural, PalmiraNeural, FiammaNeural, IrmaNeural, BenignoNeural, CataldoNeural, LisandroNeural, GianniNeural, CalimeroNeural, RinaldoNeural
  • pt-BR voices:  DonatoNeural, FabioNeural, JulioNeural, NicolauNeural, ValerioNeural, LeticiaNeural, BrendaNeural, ElzaNeural, ManuelaNeural, GiovannaNeural, LeilaNeural, YaraNeural, HumbertoNeural

 

Neural TTS and Responsible AI

We are excited about the future of Neural TTS with human like, diverse and delightful quality under the high level architecture of XYZ-Code AI framework. Our technology advancements are also guided by Microsoft’s Responsible AI process, and our principles of fairness, inclusiveness, reliability & safety, transparency, privacy & security, and accountability. We put these ethical standards into practice through the Office of Responsible AI (ORA), which sets our rules and governance processes, the AI, Ethics, and Effects in Engineering and Research (Aether) Committee, which advises our leadership on the challenges and opportunities presented by AI innovations, and Responsible AI Strategy in Engineering (RAISE), a team that enables the implementation of Microsoft responsible AI rules across engineering groups.

 

Get started

Azure AI Neural TTS offers over 440 neural voices across over 140 languages and locales. In addition, the Custom Neural Voice capability enables organizations to create a unique brand voice in multiple languages and styles.

 

For more information

 

10 Comments
Version history
Last update:
‎Dec 18 2022 09:19 PM
Updated by: