Unity使用sherpa-onnx实现离线语音合成

地狱为王

已于 2024-05-15 15:43:29 修改

阅读量235

点赞数 1

分类专栏： Unity C# 文章标签： unity 游戏引擎语音合成离线语音合成 sherpa-onnx

于 2024-05-14 16:14:45 首次发布

本文链接：https://blog.csdn.net/AWNUXCVBN/article/details/138859179

版权

Unity 同时被 2 个专栏收录

172 篇文章 10 订阅

订阅专栏

12 篇文章 0 订阅

订阅专栏

sherpa-onnx https://github.com/k2-fsa/sherpa-onnx
相关dll和lib库拷进Unity，官方示例代码稍作修改
在这里插入图片描述

using SherpaOnnx;
using System;
using System.IO;
using System.Runtime.InteropServices;
using UnityEngine;

public class TTS : MonoBehaviour
{
    public string Model = "vits-zh-aishell3/vits-aishell3.onnx";
    public string Lexicon = "vits-zh-aishell3/lexicon.txt";
    public string Tokens = "vits-zh-aishell3/tokens.txt";
    public string DataDir = "";
    public string DictDir = "";
    public string RuleFsts = "vits-zh-aishell3/rule.far";
    public int SpeakerId = 99;

    // Start is called before the first frame update
    void Start()
    {
        try
        {
            OfflineTtsConfig config = new OfflineTtsConfig();
            config.Model.Vits.Model = Path.Combine(Application.streamingAssetsPath, Model);
            config.Model.Vits.Lexicon = Path.Combine(Application.streamingAssetsPath, Lexicon);
            config.Model.Vits.Tokens = Path.Combine(Application.streamingAssetsPath, Tokens);
            //config.Model.Vits.DataDir = Path.Combine(Application.streamingAssetsPath, DataDir);
            //config.Model.Vits.DictDir = Path.Combine(Application.streamingAssetsPath, DictDir);
            config.Model.Vits.NoiseScale = 0.667f;
            config.Model.Vits.NoiseScaleW = 0.8f;
            config.Model.Vits.LengthScale = 1f;
            config.Model.NumThreads = 1;
            config.Model.Debug = 0;
            config.Model.Provider = "cpu";
            config.RuleFsts = Application.streamingAssetsPath + "/vits-zh-aishell3/phone.fst" + ","
        + Application.streamingAssetsPath + "/vits-zh-aishell3/date.fst" + ","
        + Application.streamingAssetsPath + "/vits-zh-aishell3/number.fst";
            config.MaxNumSentences = 1;
            OfflineTts tts = new OfflineTts(config);
            OfflineTtsCallback callback = new OfflineTtsCallback(MyCallback);
            OfflineTtsGeneratedAudio audio = tts.GenerateWithCallback("这是一个语音合成测试", 1, SpeakerId, callback);
            bool ok = audio.SaveToWaveFile(Application.streamingAssetsPath + "/test.wav");
            if (ok)
            {
                Debug.Log("succeeded!");
            }
            else
            {
                Debug.Log("Failed");
            }
        }
        catch (Exception e)
        {
            Debug.LogError(e);
        }
    }

    void MyCallback(IntPtr samples, int n)
    {
        float[] data = new float[n];

        Marshal.Copy(samples, data, 0, n);

        //Debug.Log("n:" + n);
    }

    // Update is called once per frame
    void Update()
    {

    }
}

在这里插入图片描述

不知为何生成的音频是8000HZ的……
24.5.15添加了方便测试的UI