Avalonia跨平台入门第四十一篇之离线语音合成

最近接到一需求:离线语音播放异常信息;听到他突然想到了去年玩耍的离线语音识别;这次还是基于讯飞再来玩耍一下,先来SDK下载:

1、找到可借鉴的代码:

https://www.cnblogs.com/DriftingLeaf/p/17648569.html

2、Windows下配置,编码格式GB2312:

string loginParams = "appid=应用的appid,work_dir=.";
string sessionBeginParams = $"engine_type=local,voice_name={speaker},text_encoding=GB2312,tts_res_path=fo|res\\tts\\{speaker}.jet;fo|res\\tts\\common.jet,sample_rate=16000,speed={speed},volume={volume},pitch=50,rdn=2,reg=2";
ret = MSPLogin(UserName, Password, loginParams);
if (ret != 0)
{
    Console.WriteLine($"MSPLogin failed, error code: {ret}.");
    return;
}
Console.WriteLine("开始合成 ...");
ret = StartSpeechSynthesis(text, filename, sessionBeginParams);
if (ret != 0)
{
    Console.WriteLine($"text_to_speech failed, error code: {ret}.");
}
Console.WriteLine("合成完毕");
MSPLogout();

3、开始合成实现方法:

private int StartSpeechSynthesis(string srcText, string desPath, string paramsStr)
{
    int ret = -1;
    nint sessionID = nint.Zero;
    uint audioLen = 0;
    WavePcmHdr wavHdr = DefaultWavHdr;
    int synthStatus = 2; // MSP_TTS_FLAG_STILL_HAVE_DATA
    if (string.IsNullOrEmpty(srcText) || string.IsNullOrEmpty(desPath))
    {
        Console.WriteLine("params is error!");
        return ret;
    }
    using (FileStream fs = new FileStream(desPath, FileMode.Create, FileAccess.Write))
    {
        sessionID = QTTSSessionBegin(paramsStr, ref ret);
        if (ret != 0)
        {
            Console.WriteLine($"QTTSSessionBegin failed, error code: {ret}.");
            return ret;
        }
        Encoding encoding = Encoding.GetEncoding("UTF-8");
        uint textLen = (uint)encoding.GetByteCount(srcText);
        ret = QTTSTextPut(sessionID, srcText, textLen, paramsStr);
        if (ret != 0)
        {
            Console.WriteLine($"QTTSTextPut failed, error code: {ret}.");
            QTTSSessionEnd(sessionID, "TextPutError");
            return ret;
        }
        Console.WriteLine("正在合成 ...");
        byte[] wavHdrBytes = StructToBytes(wavHdr);
        fs.Write(wavHdrBytes, 0, wavHdrBytes.Length);
        while (true)
        {
            nint dataPtr = QTTSAudioGet(sessionID, ref audioLen, ref synthStatus, ref ret);
            if (ret != 0)
                break;
            if (dataPtr != nint.Zero)
            {
                byte[] data = new byte[audioLen];
                Marshal.Copy(dataPtr, data, 0, (int)audioLen);
                fs.Write(data, 0, data.Length);
                wavHdr.data_size += (int)audioLen;
            }
            if (synthStatus == 2 || ret != 0)
                break;
        }
        if (ret != 0)
        {
            Console.WriteLine($"QTTSAudioGet failed, error code: {ret}.");
            QTTSSessionEnd(sessionID, "AudioGetError");
            return ret;
        }
        wavHdr.size_8 += wavHdr.data_size + (Marshal.SizeOf(wavHdr) - 8);
        fs.Seek(4, SeekOrigin.Begin);
        fs.Write(BitConverter.GetBytes(wavHdr.size_8), 0, sizeof(int));
        fs.Seek(40, SeekOrigin.Begin);
        fs.Write(BitConverter.GetBytes(wavHdr.data_size), 0, sizeof(int));
    }
    ret = QTTSSessionEnd(sessionID, "Normal");
    if (ret != 0)
    {
        Console.WriteLine($"QTTSSessionEnd failed, error code: {ret}.");
    }
    return ret;
}

4、在linux下编码格式UTF-8：

sessionBeginParams = $"engine_type = purextts,voice_name = {speaker},text_encoding =UTF-8,tts_res_path = fo|res/xtts/{speaker}.jet;fo|res/xtts/common.jet, sample_rate = 16000, speed = {speed}, volume = {volume}, pitch = 50, rdn = 2,reg=2";

5、在linux下合成实现方法:

Encoding encoding = Encoding.GetEncoding("UTF-8");
uint textLen = (uint)encoding.GetByteCount(srcText);

6、在乌班图下出现问题:

7、需要把下载SDK中资源放到输出目录下:

最终简单的效果先这样吧;以后有时间的话,可以再去摸索一下更复杂的效果;编程不息、Bug不止、无Bug、无生活;改bug的冷静、编码的激情、完成后的喜悦、挖坑的激动 、填坑的兴奋;这也许就是屌丝程序员的乐趣吧;今天就到这里吧;希望自己有动力一步一步坚持下去;生命不息,代码不止;大家抽空可以看看今天分享的效果,有好的意见和想法,可以在留言板随意留言;我看到后会第一时间回复大家,多谢大家的一直默默的关注和支持!如果觉得不错,那就伸出您的小手点个赞并关注一下!