基于C# 百度AI和科大汛飞语音合成SDK

一、百度语音合成

百度语音合成C#SDK主要是基于Rest API,需要互联网调用HTTP接口，Rest API仅支持最多512个汉字，合成的格式文件为MP3，没有其它的格式。如果想离线使用需下载SDK,Android或IOS。

1、安装语音合成 C# SDK

C# SDK 现已开源! /Baidu-AIP/dotnet-sdk

** 支持平台：.Net Framework 3.5 4.0 4.5， .Net Core 2.0 **

2、方法一：使用Nuget管理依赖 （推荐）

在NuGet中搜索Baidu.AI，安装最新版即可。

packet地址 /packages/Baidu.AI/

3、源程序界面及代码

#region 百度语音private void simpleButton1_Click(object sender, EventArgs e){if (spinEdit1.Value <= 0){spinEdit1.Focus();return;}if (trackBarControl1.Value <= 0){trackBarControl1.Focus();return;}if (string.IsNullOrEmpty(textBox1.Text.Trim())){textBox1.Focus();textBox1.Select();return;}// 设置APPID/AK/SKvar APP_ID = "******";var API_KEY = "******";var SECRET_KEY = "*****";var client = new Baidu.Aip.Speech.Tts(API_KEY, SECRET_KEY);client.Timeout = 60000; // 修改超时时间// 可选参数var option = new Dictionary<string, object>(){{"spd", spinEdit1.Value}, // 语速{"vol", trackBarControl1.Value}, // 音量{"per", comboBoxEdit1.SelectedIndex} // 发音人，4：情感度丫丫童声 };var result = client.Synthesis(textBox1.Text, option);if (xtraSaveFileDialog1.ShowDialog() == System.Windows.Forms.DialogResult.OK){if (result.ErrorCode == 0) // 或 result.Success{File.WriteAllBytes(xtraSaveFileDialog1.FileName, result.Data);}}}#endregion

接口参数说明：

二、科大讯飞语音合成

科大讯飞没有c#SDK，采用WebAPi的形式调用。不过请注意该接口使用的HTTP API协议不支持跨域。

1、接口调用流程

注：调用接口前需配置IP白名单，IP白名单规则请参照IP白名单。（由于我之前没有设置正确的IP，导致接口调用不成功）可以在百度里面输入IP将会显示你的互联网IP

通过接口密钥基于MD5计算签名，将签名以及其他参数放在Http Request Header中。将文本数据放在Http Request Body中。向服务器端发送Http请求后，接收服务器端的返回结果。

注：在控制台添加服务后，点击“发音人管理”可自行添加并试用发音人，添加后会显示该发音人参数值，设置参数voice_name为相应的发音人参数值即可。

2、程序界面及源代码

源代码

public class Rootobject{public string auf { get; set; }public string aue { get; set; }public string voice_name { get; set; }public string speed { get; set; }public string volume { get; set; }public string pitch { get; set; }public string engine_type { get; set; }public string text_type { get; set; }}

String Md5(string s){System.Security.Cryptography.MD5 md5 = new System.Security.Cryptography.MD5CryptoServiceProvider();byte[] bytes = System.Text.Encoding.UTF8.GetBytes(s);bytes = puteHash(bytes);md5.Clear();string ret = "";for (int i = 0; i < bytes.Length; i++){ret += Convert.ToString(bytes[i], 16).PadLeft(2, '0');}return ret.PadLeft(32, '0');}

#region 把流转换成缓存流MemoryStream StreamToMemoryStream(Stream instream){MemoryStream outstream = new MemoryStream();const int bufferLen = 4096;byte[] buffer = new byte[bufferLen];int count = 0;while ((count = instream.Read(buffer, 0, bufferLen)) > 0){outstream.Write(buffer, 0, count);}return outstream;}#endregion

#region 把缓存流转换成字节组public static byte[] streamTobyte(MemoryStream memoryStream){byte[] buffer = new byte[memoryStream.Length];memoryStream.Seek(0, SeekOrigin.Begin);memoryStream.Read(buffer, 0, buffer.Length);return buffer;}#endregion

private void simpleButton2_Click(object sender, EventArgs e){// 应用APPID（必须为webapi类型应用，并开通语音合成服务，参考帖子如何创建一个webapi应用：/forum.php?mod=viewthread&tid=36481）string appID = "****";// 接口密钥（webapi类型应用开通合成服务后，控制台--我的应用---语音合成---相应服务的apikey）string APIKey = "****";// 语音合成webapi接口地址String url = "/v1/service/v1/tts";String bodys;// 待合成文本string text = memoEdit1.Text;// 对要合成语音的文字先用utf-8然后进行URL加密byte[] textData = Encoding.UTF8.GetBytes(text);text = HttpUtility.UrlEncode(textData);bodys = string.Format("text={0}", text);//aue = raw, 音频文件保存类型为 wav或者pcm//aue = lame, 音频文件保存类型为 mp3string AUE = "lame";Rootobject root = new Rootobject();root.aue = AUE;root.auf = "audio/L16;rate=16000";root.speed = speed.Value.ToString();root.pitch = pitch.Value.ToString();root.volume = volume.Value.ToString();root.voice_name = voice_name.Text.Split('-')[0];root.engine_type = engine_type.Text.Split('-')[0];root.text_type = "text";string param = Newtonsoft.Json.JsonConvert.SerializeObject(root);// 获取十位的时间戳TimeSpan ts = DateTime.UtcNow - new DateTime(1970, 1, 1, 0, 0, 0, 0);string curTime = Convert.ToInt64(ts.TotalSeconds).ToString();// 对参数先utf-8然后用base64编码byte[] paramData = Encoding.UTF8.GetBytes(param);string paraBase64 = Convert.ToBase64String(paramData);// 形成签名string checkSum = Md5(APIKey + curTime + paraBase64);// 组装http请求头HttpWebRequest request = (HttpWebRequest)WebRequest.Create(url);request = (HttpWebRequest)WebRequest.Create(url);request.Method = "POST";request.ContentType = "application/x-www-form-urlencoded";request.Headers.Add("X-Param", paraBase64);request.Headers.Add("X-CurTime", curTime);request.Headers.Add("X-Appid", appID);request.Headers.Add("X-CheckSum", checkSum);Stream requestStream = request.GetRequestStream();StreamWriter streamWriter = new StreamWriter(requestStream, Encoding.GetEncoding("gb2312"));streamWriter.Write(bodys);streamWriter.Close();String htmlStr = string.Empty;HttpWebResponse response = request.GetResponse() as HttpWebResponse;Stream responseStream = response.GetResponseStream();using (StreamReader reader = new StreamReader(responseStream, Encoding.GetEncoding("UTF-8"))){string header_type = response.Headers["Content-Type"];if (header_type == "audio/mpeg"){Stream st = response.GetResponseStream();MemoryStream memoryStream = StreamToMemoryStream(st);if (xtraSaveFileDialog2.ShowDialog() == System.Windows.Forms.DialogResult.OK){// 保存音频文件地址和音频格式类型 File.WriteAllBytes(xtraSaveFileDialog2.FileName, streamTobyte(memoryStream));}Console.WriteLine(response.Headers);Console.ReadLine();}else{htmlStr = reader.ReadToEnd();Console.WriteLine(htmlStr);Console.ReadLine();}}responseStream.Close();}#endregion}

通过尝试百度和科大讯飞两个语音合成接口，发现科大讯飞语音比较好一些，百度有些词语会读错。