语音服务的语言和语音支持Language and voice support for the Speech service

语言支持因语音服务功能而异。Language support varies by Speech service functionality. 下表汇总了对语音转文本文本转语音语音翻译服务产品的语言支持。The following tables summarize language support for Speech-to-text, Text-to-speech, and Speech translation service offerings.

语音转文本Speech-to-text

Microsoft 语音 SDK 和 REST API 都支持以下语言(区域设置)。Both the Microsoft Speech SDK and the REST API support the following languages (locales).

为了提高准确性,已为一部分语言提供了自定义功能,你可通过上传音频和人工标记的脚本或相关文本(语句)进行自定义。To improve accuracy, customization is offered for a subset of the languages through uploading Audio + Human-labeled Transcripts or Related Text: Sentences. 若要了解有关自定义的详细信息,请参阅自定义语音识别入门To learn more about customization, see Get started with Custom Speech.

语言Language 区域设置 (BCP-47)Locale (BCP-47) 自定义Customizations
阿拉伯语(巴林),现代标准Arabic (Bahrain), modern standard ar-BH 语言模型Language model
阿拉伯语(埃及)Arabic (Egypt) ar-EG 语言模型Language model
阿拉伯语(伊拉克)Arabic (Iraq) ar-IQ 语言模型Language model
阿拉伯语(约旦)Arabic (Jordan) ar-JO 语言模型Language model
阿拉伯语(科威特)Arabic (Kuwait) ar-KW 语言模型Language model
阿拉伯语(黎巴嫩)Arabic (Lebanon) ar-LB 语言模型Language model
阿拉伯语(阿曼)Arabic (Oman) ar-OM 语言模型Language model
阿拉伯语(卡塔尔)Arabic (Qatar) ar-QA 语言模型Language model
阿拉伯语(沙特阿拉伯)Arabic (Saudi Arabia) ar-SA 语言模型Language model
阿拉伯语(叙利亚)Arabic (Syria) ar-SY 语言模型Language model
阿拉伯语(阿拉伯联合酋长国)Arabic (United Arab Emirates) ar-AE 语言模型Language model
中文(粤语,繁体)Chinese (Cantonese, Traditional) zh-HK 语言模型Language model
中文(普通话,简体)Chinese (Mandarin, Simplified) zh-CN 声学模型Acoustic model
语言模型Language model
中文(台湾普通话)Chinese (Taiwanese Mandarin) zh-TW 语言模型Language model
英语(澳大利亚)English (Australia) en-AU 声学模型Acoustic model
语言模型Language model
英语(加拿大)English (Canada) en-CA 声学模型Acoustic model
语言模型Language model
英语(香港)English (Hong Kong) en-HK 语言模型Language Model
英语(爱尔兰)English (Ireland) en-IE 语言模型Language Model
英语(新西兰)English (New Zealand) en-NZ 声学模型Acoustic model
语言模型Language model
英语(菲律宾)English (Philippines) en-PH 语言模型Language Model
英语(新加坡)English (Singapore) en-SG 语言模型Language Model
英语(南非)English (South Africa) en-ZA 语言模型Language Model
英语(美国)English (United States) en-US 声学模型Acoustic model
语言模型Language model
发音Pronunciation
西班牙语(阿根廷)Spanish (Argentina) es-AR 语言模型Language Model
西班牙语(玻利维亚)Spanish (Bolivia) es-BO 语言模型Language Model
西班牙语(智利)Spanish (Chile) es-CL 语言模型Language Model
西班牙语(哥伦比亚)Spanish (Colombia) es-CO 语言模型Language Model
西班牙语(哥斯达黎加)Spanish (Costa Rica) es-CR 语言模型Language Model
西班牙语(古巴)Spanish (Cuba) es-CU 语言模型Language Model
西班牙语(多米尼加共和国)Spanish (Dominican Republic) es-DO 语言模型Language Model
西班牙语(厄瓜多尔)Spanish (Ecuador) es-EC 语言模型Language Model
西班牙语(萨尔瓦多)Spanish (El Salvador) es-SV 语言模型Language Model
西班牙语(危地马拉)Spanish (Guatemala) es-GT 语言模型Language Model
西班牙语(洪都拉斯)Spanish (Honduras) es-HN 语言模型Language Model
西班牙语(墨西哥)Spanish (Mexico) es-MX 声学模型Acoustic model
语言模型Language model
西班牙(尼加拉瓜)Spanish (Nicaragua) es-NI 语言模型Language Model
西班牙语(巴拿马)Spanish (Panama) es-PA 语言模型Language Model
西班牙语(巴拉圭)Spanish (Paraguay) es-PY 语言模型Language Model
西班牙语(秘鲁)Spanish (Peru) es-PE 语言模型Language Model
西班牙语(波多黎各)Spanish (Puerto Rico) es-PR 语言模型Language Model
西班牙语(西班牙)Spanish (Spain) es-ES 声学模型Acoustic model
语言模型Language model
西班牙语(乌拉圭)Spanish (Uruguay) es-UY 语言模型Language Model
西班牙语(美国)Spanish (USA) es-US 语言模型Language Model
西班牙语(委内瑞拉)Spanish (Venezuela) es-VE 语言模型Language Model

文本转语音Text-to-speech

Microsoft 语音 SDK 和 REST API 支持以下语音,其中的每种语音都支持特定语言和方言(按区域设置标识)。Both the Microsoft Speech SDK and REST APIs support these voices, each of which supports a specific language and dialect, identified by locale. 还可以通过语音/列表 API 获取每个特定区域/终结点支持的语言和语音的完整列表。You can also get a full list of languages and voices supported for each specific region/endpoint through the voices/list API.

重要

标准语音和神经语音的定价各不相同。Pricing varies for standard and neural voices. 有关其他信息,请访问定价页。Please visit the Pricing page for additional information.

神经语音Neural voices

神经文本到语音转换是由深度神经网络提供支持的新型语音合成。Neural text-to-speech is a new type of speech synthesis powered by deep neural networks. 使用神经语音时,几乎无法将合成的语音与人类录音区分开来。When using a neural voice, synthesized speech is nearly indistinguishable from the human recordings.

使用神经语音可使得与聊天机器人和语音助手的交互更加自然且富有吸引力、将数字文本(如电子书)转换为有声读物以及增强车载导航系统。Neural voices can be used to make interactions with chatbots and voice assistants more natural and engaging, convert digital texts such as e-books into audiobooks and enhance in-car navigation systems. 随着类人的自然韵律和字词的清晰发音,用户在与 AI 系统交互时,神经语音显著减轻了听力疲劳。With the human-like natural prosody and clear articulation of words, neural voices significantly reduce listening fatigue when users interact with AI systems.

有关区域可用性的详细信息,请参阅区域For more information about regional availability, see regions.

语言Language 区域设置 (BCP-47)Locale (BCP-47) 性别Gender 语音名称Voice name 风格支持Style support
粤语(繁体中文,香港)Cantonese (Traditional Chinese, Hong Kong) zh-HK Female zh-HK-HiuGaaiNeural 常规General
英语(澳大利亚)English (Australia) en-AU FemaleFemale en-AU-NatashaNeural 常规General
英语(加拿大)English (Canada) en-CA FemaleFemale en-CA-ClaraNeural 常规General
英语(英国)English (United Kingdom) en-GB Female en-GB-LibbyNeural 常规General
英语(英国)English (United Kingdom) en-GB Female en-GB-MiaNeural 常规General
英语(英国)English (United Kingdom) en-GB 新建en-GB New Male En-GB-RyanNeural 常规General
英语(美国)English (United States) en-US Female en-US-AriaNeural 常规,提供了多种语音风格General, multiple voice styles available
英语(美国)English (United States) en-US Male en-US-GuyNeural 常规General
英语(美国)English (United States) en-US 新建en-US New Female en-US-JennyNeural 常规,提供了多种语音风格General, multiple voice styles available
普通话(简体中文,中国)Mandarin (Simplified Chinese, China) zh-CN Female zh-CN-XiaoxiaoNeural 常规,提供了多种语音风格General, multiple voice styles available
普通话(简体中文,中国)Mandarin (Simplified Chinese, China) zh-CN Female zh-CN-XiaoyouNeural 儿童语音,针对讲故事进行了优化Kid voice, optimized for story narrating
普通话(简体中文,中国)Mandarin (Simplified Chinese, China) zh-CN Male zh-CN-YunyangNeural 针对新闻播报进行了优化,提供了多种语音风格Optimized for news reading, multiple voice styles available
普通话(简体中文,中国)Mandarin (Simplified Chinese, China) zh-CN Male zh-CN-YunyeNeural 针对讲故事进行了优化Optimized for story narrating
西班牙语(墨西哥)Spanish (Mexico) es-MX FemaleFemale es-MX-DaliaNeural 常规General
西班牙语(墨西哥)Spanish (Mexico) es-MX 新建es-MX New Male es-MX-JorgeNeural 常规General
西班牙语(西班牙)Spanish (Spain) es-ES FemaleFemale es-ES-ElviraNeural 常规General
西班牙语(西班牙)Spanish (Spain) es-ES 新建es-ES New Male es-ES-AlvaroNeural 常规General

重要

en-US-JessaNeural 语音已更改为 en-US-AriaNeuralThe en-US-JessaNeural voice has changed to en-US-AriaNeural. 如果以前使用了“Jessa”,请转换为“Aria”。If you were using "Jessa" before, convert over to "Aria".

若要了解如何配置和调整神经语音,请参阅语音合成标记语言To learn how you can configure and adjust neural voices, see Speech synthesis markup language.

提示

可以继续在语音合成请求中使用完整的服务名称映射,如“Microsoft Server Speech Text to Speech Voice (en-US, AriaNeural)”。You can continue to use the full service name mapping like "Microsoft Server Speech Text to Speech Voice (en-US, AriaNeural)" in your speech synthesis requests.

标准语音Standard voices

40 多种标准语音在 10 多种语言和区域设置中提供,允许你将文本转换为合成语音。More than 40+ standard voices are available in over 10 languages and locales, which allow you to convert text into synthesized speech. 有关区域可用性的详细信息,请参阅区域For more information about regional availability, see regions.

语言Language 区域设置 (BCP-47)Locale (BCP-47) 性别Gender 语音名称Voice name
阿拉伯语(阿拉伯)Arabic (Arabic ) ar-EG FemaleFemale ar-EG-Hoda
阿拉伯语(沙特阿拉伯)Arabic (Saudi Arabia) ar-SA Male ar-SA-Naayf
粤语(繁体中文,香港)Cantonese (Traditional Chinese, Hong Kong) zh-HK Male zh-HK-Danny
粤语(繁体中文,香港)Cantonese (Traditional Chinese, Hong Kong) zh-HK FemaleFemale zh-HK-TracyRUS
克罗地亚语(克罗地亚)Croatian (Croatia) hr-HR Male hr-HR-Matej
英语(澳大利亚)English (Australia) en-AU FemaleFemale en-AU-Catherine
英语(澳大利亚)English (Australia) en-AU FemaleFemale en-AU-HayleyRUS
英语(加拿大)English (Canada) en-CA FemaleFemale en-CA-HeatherRUS
英语(加拿大)English (Canada) en-CA FemaleFemale en-CA-Linda
英语(英国)English (United Kingdom) en-GB Male en-GB-George
英语(英国)English (United Kingdom) en-GB FemaleFemale en-GB-HazelRUS
英语(英国)English (United Kingdom) en-GB FemaleFemale en-GB-Susan
英语(美国)English (United States) en-US Male en-US-BenjaminRUS
英语(美国)English (United States) en-US Male en-US-GuyRUS
英语(美国)English (United States) en-US FemaleFemale en-US-JessaRUS
英语(美国)English (United States) en-US FemaleFemale en-US-ZiraRUS
法语(加拿大)French (Canada) fr-CA FemaleFemale fr-CA-Caroline
法语(加拿大)French (Canada) fr-CA FemaleFemale fr-CA-HarmonieRUS
法语(瑞士)French (Switzerland) fr-CH Male fr-CH-Guillaume
德语(奥地利)German (Austria) de-AT Male de-AT-Michael
德语(德国)German (Germany) de-DE Male de-DE-Stefan
印地语(印度)Hindi (India) hi-IN Male hi-IN-Hemant
印地语(印度)Hindi (India) hi-IN FemaleFemale hi-IN-Kalpana
意大利语(意大利)Italian (Italy) it-IT Male it-IT-Cosimo
意大利语(意大利)Italian (Italy) it-IT FemaleFemale it-IT-LuciaRUS
韩语(韩国)Korean (Korea) ko-KR FemaleFemale ko-KR-HeamiRUS
普通话(简体中文,中国)Mandarin (Simplified Chinese, China) zh-CN FemaleFemale zh-CN-HuihuiRUS
普通话(简体中文,中国)Mandarin (Simplified Chinese, China) zh-CN Male zh-CN-Kangkang
普通话(简体中文,中国)Mandarin (Simplified Chinese, China) zh-CN FemaleFemale zh-CN-Yaoyao
普通话(繁体中文,台湾)Mandarin (Traditional Chinese, Taiwan) zh-TW FemaleFemale zh-TW-HanHanRUS
普通话(繁体中文,台湾)Mandarin (Traditional Chinese, Taiwan) zh-TW FemaleFemale zh-TW-Yating
普通话(繁体中文,台湾)Mandarin (Traditional Chinese, Taiwan) zh-TW Male zh-TW-Zhiwei
俄语(俄罗斯)Russian (Russia) ru-RU FemaleFemale ru-RU-EkaterinaRUS
俄语(俄罗斯)Russian (Russia) ru-RU FemaleFemale ru-RU-Irina
俄语(俄罗斯)Russian (Russia) ru-RU Male ru-RU-Pavel
西班牙语(墨西哥)Spanish (Mexico) es-MX FemaleFemale es-MX-HildaRUS
西班牙语(墨西哥)Spanish (Mexico) es-MX Male es-MX-Raul
西班牙语(西班牙)Spanish (Spain) es-ES FemaleFemale es-ES-HelenaRUS
西班牙语(西班牙)Spanish (Spain) es-ES FemaleFemale es-ES-Laura
西班牙语(西班牙)Spanish (Spain) es-ES Male es-ES-Pablo

重要

en-US-Jessa 语音已更改为 en-US-AriaThe en-US-Jessa voice has changed to en-US-Aria. 如果以前使用了“Jessa”,请转换为“Aria”。If you were using "Jessa" before, convert over to "Aria".

提示

可以继续在语音合成请求中使用完整的服务名称映射,如“Microsoft Server Speech Text to Speech Voice (en-US, AriaRUS)”。You can continue to use the full service name mapping like "Microsoft Server Speech Text to Speech Voice (en-US, AriaRUS)" in your speech synthesis requests.

自定义Customization

语音自定义适用于 en-GBen-INen-USes-MXzh-CNVoice customization is available for en-GB, en-IN, en-US, es-MX and zh-CN. 选择与训练自定义语音模型所需的训练数据相匹配的正确区域设置。Select the right locale that matches the training data you have to train a custom voice model. 例如,如果你的录音数据是以带英国口音的英语说出的,请选择 en-GBFor example, if the recording data you have is spoken in English with a British accent, select en-GB.

备注

除了中英双语模型之外,我们在自定义语音中不支持其他双语模型训练。We do not support bi-lingual model training in Custom Voice, except for the Chinese-English bi-lingual. 如果要训练一种也可以说英语的中文语音,请选择“中英双语”。Select "Chinese-English bilingual" if you want to train a Chinese voice that can speak English as well. 对于 en-USzh-CN 之外的所有区域设置,语音训练都从一个包含 2000 条以上言语的数据集开始。对于这例外的两种区域设置,你可以从任何大小的训练数据开始。Voice training in all locales starts with a data set of 2,000+ utterances, except for the en-US and zh-CN where you can start with any size of training data.

语音翻译Speech translation

语音翻译 API 支持使用不同的语言进行语音转语音和语音转文本的翻译。The Speech Translation API supports different languages for speech-to-speech and speech-to-text translation. 源语言必须始终来自“语音转文本”语言表。The source language must always be from the Speech-to-text language table. 可用的目标语言取决于翻译目标是语音还是文本。The available target languages depend on whether the translation target is speech or text. 可以将传入的语音翻译成 60 种以上的语言You may translate incoming speech into more than 60 languages. 这些语言的子集可用于语音合成A subset of languages are available for speech synthesis.

文本语言Text languages

文本语言Text language 语言代码Language code
阿拉伯语Arabic ar
简体中文Chinese Simplified zh-Hans
中文(繁体)Chinese Traditional zh-Hant
英语English en
法语French fr
德语German de
HindiHindi hi
朝鲜语Korean ko
俄语Russian ru
西班牙语Spanish es

更多语言More languages

备注

如果前面的列表中不支持你指定的语言,请参考主权云支持的区域设置联系支持,可以根据要求部署该语言。If the language you specified is not supported in the previous list, please refer to sovereign clouds supported locales and contact support, the language could be deployed on your request.

后续步骤Next steps