400 898 2016
产品简介
基本概念
平台新手指引
计价模式
获取访问令牌
语音合成
语音识别
声音复刻
声音转换
离线声音转换
声纹识别
声音理解
协议规则
常见问题语音合成常见问题
xml标签如何使用
音子边界信息（interval时间戳）
系统收到一段文本的合成请求后，会进行文本分析、韵律分析生产音子序列在进行音频合成，每合成一个小片段系统返回一次音频（包括静音），返回音频后计算音频的音子边界信息， 音子边界返回的时间单位是秒，每个小句的音频会从头开始计时
-
音子合集
"SIL","SP3",
"ZH","Z","Y","W","V","UW","UHR","UH","TS","TR","TH","T","SH","S","R","P","OY","OW","NG","N","M","L","K","JH","IY","IHR","IH","HH","G","F","EY","ER","EHR","EH","DZ","DR","DH","D","CH","B","AY","AX","AW","AO","AH","AE","AA",
"b","ch","c","d","f","g","h","j","k","l","m","n","p","q","r","sh","s","t","x","zh","z",
"a","ai","an","ang","ao","e","ei","en","eng","er","i","ia","ian","iang","iao","ie","in","ing","io","iong","iou","iu","o","ong","ou","u","ua","uai","uan","uang","uei","uen","ueng","ui","un","uo","v","van","ve","vn",
"air","angr","anr","aor","ar","eir","engr","enr","iangr","ianr","iaor","iar","ier","ingr","inr","ir","iur","ongr","or","our","uair","uangr","uanr","uar","ueir","uengr","uenr","uir","unr","uor","ur","vanr","ver","vnr","vr"
SIL和SP3都表示静音，其他大写字母为英文音标，小写字母为中文的音素。
- 国际音标表
| 国际音标 | CMU | 例子 | 国际音标 | CMU | 例子 | 国际音标 | CMU | 例子 |
|---|---|---|---|---|---|---|---|---|
| ɑ: | 美音不存在，会读成ae,ar(若有r时） | au | AW | cow | y | Y | yield | |
| ʌ | AH | hut | əu | OW | no | θ | TH | theta |
| ə | AX | |||||||
| ɔ/ɑ | AA | odd | uə | UH R | poor | ð | DH | the |
| ɔ: | AO | ought | ɛə | EH R | there | ʃ | SH | she |
| ɝ | ER | hurt | p | P | pee | ʒ | ZH | seizure |
| ə: | 美音不存在 | 美音不存在 | b | B | be | tʃ | CH | cheese |
| ı | IH | it | t | T | tea | dʒ | JH | gee |
| i: | IY | eat | d | D | dee | tr | T R | tree |
| u | UH | hood | k | K | key | dr | D R | dress |
| u: | UW | who | g | G | green | ts | T S | effects |
| e | EH | ed | m | M | me | dz | D Z | fields |
| æ | AE | at | n | N | knee | v | V | vee |
| aı | AY | hide | ŋ | NG | ping | w | W | we |
| eı | EY | detail | l | L | lee | f | F | fee |
| ɔı | OY | toy | s | S | sea | r | R | read |
| ıə | IH R | hear | z | Z | zee | h | HH | he |
什么是定制库
-
什么是定制库？
定制库是根据客户提供的话术文本进行指定发音人的录音定制，然后集成到客户购买的tts中，调用的时候如果合成文本包含客户定制的文本则直接取定制的录音音频返回给调用方。
-
制作定制库的流程是什么？
客户提供需要定制的文本，我们进行此文本的录音录制，再经过数据的加工生成一个.dat的文件，配置到tts中即可。
-
什么时候会考虑使用定制库？
- 高频的固定话术（比如客服场景，智能设备唤醒词）；
- 通过tts合成的效果达不到客户的要求；
- 我方短期内无法优化达到客户的要求；
如果有定制库的部分，我们会在发给你们安装包的邮件中详细说明哪些文件是定制库，如果没有说明则表示此音库没有定制库文件，也不需要相关定制库的配置。
