xuning / sherpaonnx

转到一个项目
Toggle navigation Toggle navigation pinning
Fangjun Kuang
Committed by GitHub
d9c586cc 1 parent 356da3b5

Removed unused TTS example code in .Net examples (#1492)

内嵌 并排对比
正在显示 15 个修改的文件 包含 7 行增加 和 495 行删除
... ... @@ -20,6 +20,7 @@ if [ x"$EMSCRIPTEN" == x"" ]; then
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
emcc --version
fi
fi
... ...
... ... @@ -15,6 +15,7 @@ if [ x"$EMSCRIPTEN" == x"" ]; then
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
emcc --version
fi
fi
... ...
... ... @@ -22,6 +22,7 @@ if [ x"$EMSCRIPTEN" == x"" ]; then
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
emcc --version
fi
fi
... ...
... ... @@ -20,6 +20,7 @@ if [ x"$EMSCRIPTEN" == x"" ]; then
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
emcc --version
fi
fi
... ...
... ... @@ -20,6 +20,7 @@ if [ x"$EMSCRIPTEN" == x"" ]; then
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
emcc --version
fi
fi
... ...
... ... @@ -21,6 +21,7 @@ if [ x"$EMSCRIPTEN" == x"" ]; then
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
emcc --version
fi
fi
... ...
... ... @@ -20,6 +20,7 @@ if [ x"$EMSCRIPTEN" == x"" ]; then
exit 1
else
EMSCRIPTEN=$(dirname $(realpath $(which emcc)))
emcc --version
fi
fi
... ...
using NAudio.Wave;
namespace TTS.Struct
{
public sealed partial class SherpaOnnxGeneratedAudioResult
{
private WaveOutEvent waveOut;
private WaveFormat waveFormat;
private BufferedWaveProvider bufferedWaveProvider;
private int bufferLength = 1;
public TimeSpan? AudioDuration => bufferedWaveProvider?.BufferedDuration;
public float PlayProgress => (waveOut?.GetPosition() * 1.0f / bufferLength).Value;
public void Play()
{
waveOut ??= new WaveOutEvent();
waveFormat ??= new WaveFormat(sample_rate, AudioDataBit, Channels); // 32-bit 浮点，单声道
if (bufferedWaveProvider == null)
{
bufferedWaveProvider ??= new BufferedWaveProvider(waveFormat);
var buffer = AudioByteData;
bufferLength = buffer.Length;
bufferedWaveProvider.AddSamples(buffer, 0, bufferLength);
bufferedWaveProvider.BufferLength = bufferLength;
waveOut.Init(bufferedWaveProvider);
}
waveOut.Play();
}
public void Stop()
{
waveOut?.Stop();
}
}
}
using System.Text;
using TTS;
using TTS.Struct;
internal class Program
{
private static void Main(string[] args)
{
SherpaOnnxOfflineTtsConfig sherpaOnnxOfflineTtsConfig = new SherpaOnnxOfflineTtsConfig();
sherpaOnnxOfflineTtsConfig.model = new SherpaOnnxOfflineTtsModelConfig
{
debug = 0,
num_threads = 4,
provider = "cpu",
vits = new SherpaOnnxOfflineTtsVitsModelConfig
{
//lexicon = "vits-zh-aishell3/lexicon.txt",
//model = "vits-zh-aishell3/vits-aishell3.onnx",
//tokens = "vits-zh-aishell3/tokens.txt",
model = @"C:\Services\Sherpa\model.onnx",
lexicon = "",
tokens = @"C:\Services\Sherpa\tokens.txt",
data_dir = @"C:\Services\Sherpa\espeak-ng-data",
noise_scale = 0.667f,
noise_scale_w = 0.8f,
length_scale = 1,
},
};
TTSCore i = new TTSCore(sherpaOnnxOfflineTtsConfig);
Console.InputEncoding = Encoding.Unicode;
Console.OutputEncoding = Encoding.UTF8;
while (true)
{
var str = Console.ReadLine();
var audioResult = i.ToSpeech(str, 40, 1f);
// audioResult.WriteWAVFile("123.wav");保存本地
audioResult.Play();
int lastIndex = -1;
while (audioResult.PlayProgress <= 1f)
{
int index = (int)(audioResult.PlayProgress * (str.Length - 1));
if (lastIndex != index)
{
Console.Write(str[index]);
lastIndex = index;
}
Thread.Sleep(100);
}
if (++lastIndex < str.Length)
Console.Write(str[lastIndex]);
Console.WriteLine();
}
}
}
using System;
using System.Collections.Generic;
using System.Linq;
using System.Runtime.InteropServices;
using System.Text;
using System.Threading.Tasks;
namespace TTS.Struct
{
/// <summary>
/// 生成语音结果
/// </summary>
public sealed partial class SherpaOnnxGeneratedAudioResult : IDisposable
{
public const string Filename = "sherpa-onnx-c-api";
/// <summary>
/// 销毁非托管内存
/// </summary>
/// <param name="ttsGenerateIntptr"></param>
[DllImport(Filename)]
private static extern void SherpaOnnxDestroyOfflineTtsGeneratedAudio(IntPtr ttsGenerateIntptr);
[DllImport(Filename)]
private static extern int SherpaOnnxWriteWave(IntPtr q, int n, int sample_rate, string filename);
/// <summary>
/// 音频数据比特
/// </summary>
public const int AudioDataBit = 16;
/// <summary>
/// 单通道
/// </summary>
public const int Channels = 1;
/// <summary>
/// 原生句柄
/// </summary>
internal IntPtr thisHandle;
internal readonly IntPtr audioData;
internal readonly int dataSize;
/// <summary>
/// 采样率
/// </summary>
public readonly int sample_rate;
/// <summary>
/// 音频数据指针
/// </summary>
public IntPtr AudioDataIntPtr => audioData;
/// <summary>
/// 数据的大小
/// </summary>
public unsafe int AudioDataLength
{
get
{
return dataSize;
//float* buffer = (float*)audioData;
//while (*buffer != 0)
// ++buffer;
//return (int)(buffer - (float*)audioData);
}
}
/// <summary>
/// 获得音频数据 float[]
/// 这个内部创建一个数组
/// </summary>
public unsafe float[] AudioFloatData
{
get
{
int length = AudioDataLength;
float[] floatAudioData = new float[length];
Marshal.Copy(audioData, floatAudioData, 0, floatAudioData.Length);
return floatAudioData;
}
}
/// <summary>
/// 获得音频数据 byte[]
/// 这个内部创建一个数组
/// </summary>
public byte[] AudioByteData
{
get
{
byte[] bytes = new byte[AudioDataLength * 2];
ReadData(bytes, 0);
return bytes;
}
}
internal SherpaOnnxGeneratedAudioResult(IntPtr intPtr, SherpaOnnxGeneratedAudio sherpaOnnx)
{
this.thisHandle = intPtr;
this.audioData = sherpaOnnx.audioData;
this.dataSize = sherpaOnnx.dataSize;
this.sample_rate = sherpaOnnx.sample_rate;
}
~SherpaOnnxGeneratedAudioResult()
{
Dispose();
}
/// <summary>
/// 读取数据
/// 没有垃圾产生，自己传递数组进来
/// </summary>
/// <param name="audioFloats">数组</param>
/// <param name="offset">数组那个位置写入</param>
/// <returns>写入了多少个</returns>
public int ReadData(float[] audioFloats, int offset)
{
int length = AudioDataLength;
int c = audioFloats.Length - offset;
length = c >= length ? length : c;
Marshal.Copy(audioData, audioFloats, offset, length);
return length;
}
/// <summary>
/// 读取数据
/// 这个内部转换成byte[] 音频数组
/// 没有垃圾产生，自己传递数组进来
/// </summary>
/// <param name="audioFloats">数组，这个长度需要是AudioDataLength*2大小</param>
/// <param name="offset">数组那个位置写入</param>
/// <returns>写入了多少个</returns>
public int ReadData(byte[] audioFloats, int offset)
{
//因为是16bit存储音频数据，所以float会转换成两个字节存储
var audiodata = AudioFloatData;
int length = audiodata.Length * 2;
int c = audioFloats.Length - offset;
c = c % 2 == 0 ? c : c - 1;
length = c >= length ? length : c;
int p = length / 2;
for (int i = 0; i < p; i++)
{
short value = (short)(audiodata[i] * short.MaxValue);
audioFloats[offset++] = (byte)value;
audioFloats[offset++] = (byte)(value >> 8);
}
return length;
}
/// <summary>
/// 写入WAV音频数据
/// </summary>
/// <param name="filename"></param>
/// <returns></returns>
public bool WriteWAVFile(string filename)
{
return 1 == SherpaOnnxWriteWave(audioData, this.dataSize, this.sample_rate, filename);
}
public void Dispose()
{
if (this.thisHandle != IntPtr.Zero)
{
SherpaOnnxDestroyOfflineTtsGeneratedAudio(this.thisHandle);
GC.SuppressFinalize(this);
this.thisHandle = IntPtr.Zero;
}
}
}
[StructLayout(LayoutKind.Sequential)]
internal struct SherpaOnnxGeneratedAudio
{
internal readonly IntPtr audioData;
internal readonly int dataSize;
/// <summary>
/// 采样率
/// </summary>
public readonly int sample_rate;
}
}
using System.Runtime.InteropServices;
namespace TTS.Struct
{
[StructLayout(LayoutKind.Sequential)]
public struct SherpaOnnxOfflineTtsConfig
{
public SherpaOnnxOfflineTtsModelConfig model;
[MarshalAs(UnmanagedType.LPStr)]
public string rule_fsts;
public int max_num_sentences;
[MarshalAs(UnmanagedType.LPStr)]
public string rule_fars;
}
}
using System.Runtime.InteropServices;
namespace TTS.Struct
{
[StructLayout(LayoutKind.Sequential)]
public struct SherpaOnnxOfflineTtsModelConfig
{
/// <summary>
/// 模型配置
/// </summary>
public SherpaOnnxOfflineTtsVitsModelConfig vits;
/// <summary>
/// 线程数
/// </summary>
public int num_threads;
public int debug;
/// <summary>
/// 使用cpu
/// </summary>
[MarshalAs(UnmanagedType.LPStr)]
public string provider;
}
}
using System.Runtime.InteropServices;
namespace TTS.Struct
{
[StructLayout(LayoutKind.Sequential)]
public struct SherpaOnnxOfflineTtsVitsModelConfig
{
/// <summary>
/// 模型
/// "vits-zh-aishell3/vits-aishell3.onnx"
/// </summary>
[MarshalAs(UnmanagedType.LPStr)]
public string model;
/// <summary>
/// 词典文件
/// "vits-zh-aishell3/lexicon.txt"
/// </summary>
[MarshalAs(UnmanagedType.LPStr)]
public string lexicon;
[MarshalAs(UnmanagedType.LPStr)]
public string tokens;
[MarshalAs(UnmanagedType.LPStr)]
public string data_dir;
/// <summary>
/// VITS模型的noise_scale (float，默认值= 0.667)
/// </summary>
public float noise_scale = 0.667f;
/// <summary>
/// VITS模型的noise_scale_w (float，默认值= 0.8)
/// </summary>
public float noise_scale_w = 0.8f;
/// <summary>
/// 演讲的速度。大→慢;小→更快。(float, default = 1)
/// </summary>
public float length_scale = 1f;
[MarshalAs(UnmanagedType.LPStr)]
public string dict_dir;
public SherpaOnnxOfflineTtsVitsModelConfig()
{
noise_scale = 0.667f;
noise_scale_w = 0.8f;
length_scale = 1f;
model = "vits-zh-aishell3/vits-aishell3.onnx";
lexicon = "vits-zh-aishell3/lexicon.txt";
tokens = "vits-zh-aishell3/tokens.txt";
data_dir = "";
dict_dir = "";
}
}
}
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
<TargetFramework>net6.0</TargetFramework>
<ImplicitUsings>enable</ImplicitUsings>
<Nullable>enable</Nullable>
<AllowUnsafeBlocks>true</AllowUnsafeBlocks>
</PropertyGroup>
<ItemGroup>
<PackageReference Include="NAudio" Version="2.2.1" />
<PackageReference Include="org.k2fsa.sherpa.onnx" Version="1.10.23" />
</ItemGroup>
</Project>
using System.Runtime.InteropServices;
using TTS.Struct;
namespace TTS
{
internal sealed class TTSCore : IDisposable
{
public const string Filename = "sherpa-onnx-c-api";
[DllImport(Filename, CallingConvention = CallingConvention.Cdecl)]
private static extern IntPtr SherpaOnnxCreateOfflineTts(SherpaOnnxOfflineTtsConfig handle);
[DllImport(Filename)]
private static extern IntPtr SherpaOnnxOfflineTtsGenerate(IntPtr createOfflineTtsIntptr, IntPtr text, int sid, float speed);
[DllImport(Filename)]
private static extern void SherpaOnnxDestroyOfflineTts(IntPtr intPtr);
/// <summary>
/// 原生句柄
/// </summary>
private IntPtr thisHandle;
public TTSCore(SherpaOnnxOfflineTtsConfig modelConfig)
{
IntPtr ttsHandle = SherpaOnnxCreateOfflineTts(modelConfig);
if (ttsHandle == IntPtr.Zero)
{
throw new InvalidOperationException("Failed to create SherpaOnnx TTS engine.");
}
thisHandle = ttsHandle;
}
/// <summary>
/// 文字转语音
/// </summary>
/// <param name="text">文字</param>
/// <param name="sid">音色</param>
/// <param name="speed">速度</param>
/// <returns></returns>
public SherpaOnnxGeneratedAudioResult ToSpeech(string text, int sid, float speed = 1f)
{
var result = SherpaOnnxOfflineTtsGenerate(thisHandle, Marshal.StringToCoTaskMemUTF8(text), sid, speed);
SherpaOnnxGeneratedAudio impl = (SherpaOnnxGeneratedAudio)Marshal.PtrToStructure(result, typeof(SherpaOnnxGeneratedAudio));
return new SherpaOnnxGeneratedAudioResult(result, impl);
}
/// <summary>
/// 文字转语音
/// </summary>
/// <param name="text">文字</param>
/// <param name="sid">音色</param>
/// <param name="speed">速度</param>
/// <returns></returns>
public Task<SherpaOnnxGeneratedAudioResult> ToSpeechAsync(string text, int sid, float speed = 1f)
{
return Task.Run(() => ToSpeech(text, sid, speed));
}
~TTSCore()
{
Dispose();
}
public void Dispose()
{
if (this.thisHandle != IntPtr.Zero)
{
SherpaOnnxDestroyOfflineTts(this.thisHandle);
GC.SuppressFinalize(this);
this.thisHandle = IntPtr.Zero;
}
}
}
}