【API解析】微软edge浏览器大声朗读功能（read aloud）调用步骤
1. 来源
- github: MsEdgeTTS, edge-TTS-record
- 吾爱破解：微软语音助手免费版，支持多种功能，全网首发
2. 准备工作
- 功能来源：edge浏览器
- 抓包工具：fiddler
- 模拟请求：postman
3. 主要分析步骤
- 第一步：确定edge浏览器read aloud功能用js如何调用，fiddler上没有捕捉到
const voices = speechSynthesis.getVoices()
function speakbyvoice(text, voice) {
var utter = new SpeechSynthesisUtterance(text)
for (let v of voices) {
if (v.name.includes(voice)) {
utter.voice = v
break
}
}
speechSynthesis.speak(utter)
return utter
}
speakbyvoice("hello world", "Xiaoxiao")
- 第二步：试着对edge-TTS-record抓包，抓到了一个http请求和websocket连接。对照MsEdgeTTS的代码可知：
/*
* postman中模拟成功
* 获取可用语音包选项，等价于speechSynthesis.getVoices()
* http url: https://speech.platform.bing.com/consumer/speech/synthesize/readaloud/voices/list?trustedclienttoken=6A5AA1D4EAFF4E9FB37E23D68491D6F4
* method: GET
*/
{
uri: "https://speech.platform.bing.com/consumer/speech/synthesize/readaloud/voices/list",
query: {
trustedclienttoken: "6A5AA1D4EAFF4E9FB37E23D68491D6F4"
}
method: "GET"
}
/*
* postman中模拟成功
* 发送wss连接，传输文本和语音数据，等价于speechSynthesis.speak(utter)
* wss url: wss://speech.platform.bing.com/consumer/speech/synthesize/readaloud/edge/v1?TrustedClientToken=
* send: 发送两次数据，第一次是需要的音频格式，第二次是ssml标记文本（需要随机生成一个requestid，替换掉guid的分隔符“-”即可）
* receive: 接收到的webm音频字节包含在相同requestid的正文部分，用Path=audio\r\n定位正文索引
* 存在的问题: 1、第一次发送的音频格式文本中，只有在webm-24khz-16bit-mono-opus格式下才能成功连接，其他格式尝试后直接断开；
* 2、第二次发送的ssml文本不支持mstts命名空间的解析，是Auzure语音服务的阉割版，例如不能出现xmlns:mstts="****"、<mstts:express-as/>、<p/>、<s/>等语言标记
*/
{
uri: "https://speech.platform.bing.com/consumer/speech/synthesize/readaloud/voices/list",
query: {
trustedclienttoken: "6A5AA1D4EAFF4E9FB37E23D68491D6F4"
},
sendmessage: {
audioformat: `
X-Timestamp:Mon Jul 11 2022 17:50:42 GMT+0800 (中国标准时间)
Content-Type:application/json; charset=utf-8
Path:speech.config
{"context":{"synthesis":{"audio":{"metadataoptions":{"sentenceBoundaryEnabled":"false","wordBoundaryEnabled":"true"},"outputFormat":"webm-24khz-16bit-mono-opus"}}}}`,
ssml: `
X-RequestId:7e956ecf481439a86eb1beec26b4db5a
Content-Type:application/ssml+xml
X-Timestamp:Mon Jul 11 2022 17:50:42 GMT+0800 (中国标准时间)Z
Path:ssml
<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xml:lang='en-US'><voice name='Microsoft Server Speech Text to Speech Voice (zh-CN, XiaoxiaoNeural)'><prosody pitch='+0Hz' rate ='+0%' volume='+0%'> hello world</prosody></voice></speak>`
}
}
4. 编写代码
- websocket库：WebSocketSharp。最新版安装失败的可以降版本安装，此文发布的时候最新预览版是
1.0.3-rc11
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text.RegularExpressions;
using WebSocketSharp; // nuget包：WebSocketSharp（作者：sta，此文安装版本：1.0.3-rc10）
namespace ConsoleTest
{
internal class Program
{
static string ConvertToAudioFormatWebSocketString(string outputformat)
{
return "Content-Type:application/json; charset=utf-8\r\nPath:speech.config\r\n\r\n{\"context\":{\"synthesis\":{\"audio\":{\"metadataoptions\":{\"sentenceBoundaryEnabled\":\"false\",\"wordBoundaryEnabled\":\"false\"},\"outputFormat\":\"" + outputformat + "\"}}}}";
}
static string ConvertToSsmlText(string lang, string voice, string text)
{
return $"<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xmlns:mstts='https://www.w3.org/2001/mstts' xml:lang='{
lang}'><voice name='{
voice}'>{
text}</voice></speak>";
}
static string ConvertToSsmlWebSocketString(string requestId, string lang, string voice, string msg)
{
return $"X-RequestId:{
requestId}\r\nContent-Type:application/ssml+xml\r\nPath:ssml\r\n\r\n{
ConvertToSsmlText(lang, voice, msg)}";
}
static void Main(string[] args)
{
var url = "wss://speech.platform.bing.com/consumer/speech/synthesize/readaloud/edge/v1?trustedclienttoken=6A5AA1D4EAFF4E9FB37E23D68491D6F4";
var Language = "en-US";
var Voice = "Microsoft Server Speech Text to Speech Voice (zh-CN, XiaoxiaoNeural)";
var audioOutputFormat = "webm-24khz-16bit-mono-opus";
var binary_delim = "Path:audio\r\n";
var msg = "Hello world";
var sendRequestId = Guid.NewGuid().ToString().Replace("-", "");
var dataBuffers = new Dictionary<string, List<byte>>();
var webSocket = new WebSocket(url);
webSocket.SslConfiguration.ServerCertificateValidationCallback = (sender, certificate, chain, sslPolicyErrors) => true;
webSocket.OnOpen += (sender, e) => Console.WriteLine("[Log] WebSocket Open");
webSocket.OnClose += (sender, e) => Console.WriteLine("[Log] WebSocket Close");
webSocket.OnError += (sender, e) => Console.WriteLine("[Error] error message: " + e.Message);
webSocket.OnMessage += (sender, e) =>
{
if (e.IsText)
{
var data = e.Data;
var requestId = Regex.Match(data, @"X-RequestId:(?<requestId>.*?)\r\n").Groups["requestId"].Value;
if (data.Contains("Path:turn.start"))
{
// start of turn, ignore. 开始信号，不用处理
}
else if (data.Contains("Path:turn.end"))
{
// end of turn, close stream. 结束信号，可主动关闭socket
// dataBuffers[requestId] = null;
// 不要跟着MsEdgeTTS中用上面那句，音频发送完毕后，最后还会收到一个表示音频结束的文本信息
webSocket.Close();
}
else if (data.Contains("Path:response"))
{
// context response, ignore. 响应信号，无需处理
}
else
{
Console.WriteLine("unknow message: " + data); // 未知错误，通常不会发生
}
}
else if (e.IsBinary)
{
var data = e.RawData;
var requestId = Regex.Match(e.Data, @"X-RequestId:(?<requestId>.*?)\r\n").Groups["requestId"].Value;
if (!dataBuffers.ContainsKey(requestId))
dataBuffers[requestId] = new List<byte>();
if (data[0] == 0x00 && data[1] == 0x67 && data[2] == 0x58)
{
// Last (empty) audio fragment. 空音频片段，代表音频发送结束
}
else
{
var index = e.Data.IndexOf(binary_delim) + binary_delim.Length;
dataBuffers[requestId].AddRange(data.Skip(index));
}
}
};
webSocket.Connect();
var audioconfig = ConvertToAudioFormatWebSocketString(audioOutputFormat);
webSocket.Send(audioconfig);
webSocket.Send(ConvertToSsmlWebSocketString(sendRequestId, Language, Voice, msg));
while (webSocket.IsAlive) {
}
Console.WriteLine("接收到的音频字节长度：" + dataBuffers[sendRequestId].Count);
Console.ReadKey(true);
}
}
}
5. 结语
模拟websocket请求成功，缺陷是postman模拟结果显示音频outputformat参数只能是webm-24khz-16bit-mono-opus，也就是说还需要再用ffmpeg之类的库转换格式。暂时也没找到比较好用的库，先记录到这