Unity 工具 之 Azure 微软语音合成普通方式和流式获取音频数据的简单整理_unity音频流式传输

Unity 工具 之 Azure 微软语音合成普通方式和流式获取音频数据的简单整理

目录

Unity 工具 之 Azure 微软语音合成普通方式和流式获取音频数据的简单整理

一、简单介绍

二、实现原理

三、注意实现

四、实现步骤

 六、关键脚本

附加：

声音设置相关

---

一、简单介绍

Unity 工具类，自己整理的一些游戏开发可能用到的模块，单独独立使用，方便游戏开发。

本节介绍，这里在使用微软的Azure 进行语音合成的两个方法的做简单整理，这里简单说明，如果你有更好的方法，欢迎留言交流。

官网注册：

面向学生的 Azure - 免费帐户额度 | Microsoft Azure

官网技术文档网址：

技术文档 | Microsoft Learn

官网的TTS：

文本转语音快速入门 - 语音服务 - Azure Cognitive Services | Microsoft Learn

Azure Unity SDK  包官网：

安装语音 SDK - Azure Cognitive Services | Microsoft Learn

SDK具体链接：

https://aka.ms/csspeech/unitypackage

二、实现原理

1、官网申请得到语音合成对应的 SPEECH_KEY 和 SPEECH_REGION

2、然后对应设置 语言 和需要的声音 配置

3、使用 普通方式 和 流式获取得到音频数据，在声源中播放即可

三、注意实现

1、在合成语音文本较长的情况下，流式获取的速度明显会优于普通的方式

2、目前流式获取的方式，我是暂时没有好的方式管理网络错误和音频播放结束的事件

（如果有兄弟集美们知道，还请留言赐教哈）

四、实现步骤

1、下载好SDK 导入

2、简单的搭建场景

3、写测试脚本，和普通获取和流式获取方式

4、把测试脚本添加到场景中，并赋值

 5、运行，输入文字，点击对应按钮即可

 六、关键脚本

1、Test

 
using UnityEngine;
using UnityEngine.UI;
 
public class Test : MonoBehaviour
{
    public InputField m_InputField;
    public Button m_StreamButton;
    public Button m_NormalButton;
    public AudioSource m_AudioSource;
 
    // Start is called before the first frame update
    void Start()
    {
 
        m_StreamButton.onClick.AddListener(() => {
            AzureTTSStream.Instance.StartTTS(m_InputField.text, m_AudioSource);
        });
        m_NormalButton.onClick.AddListener(() => {
            AzureTTSNormal.Instance.StartTTS(m_InputField.text, m_AudioSource);
        });
    }
 
}

2、AzureTTSNormal

using Microsoft.CognitiveServices.Speech;
using System;
using System.Collections;
using UnityEngine;
 
public class AzureTTSNormal : MonoSingleton<AzureTTSNormal>
{
    private AudioSource m_AudioSource;
 
    private string m_SubscriptionKey = "Your";
    private string m_Region = "Your";
    private string m_SpeechSynthesisLanguage = "zh-CN";
    private string m_SpeechSynthesisVoiceName = "zh-CN-XiaochenNeural";
    private Coroutine m_TTSCoroutine;
 
    /// <summary>
    /// 你的授权
    /// </summary>
    /// <param name="subscriptionKey">子脚本的Key</param>
    /// <param name="region">地区</param>
    public void SetAzureAuthorization(string subscriptionKey, string region)
    {
        m_SubscriptionKey = subscriptionKey;
        m_Region = region;
    }
 
    /// <summary>
    /// 设置语音和声音
    /// </summary>
    /// <param name="language">语言</param>
    /// <param name="voiceName">声音</param>
    public void SetLanguageVoiceName(SpeechSynthesisLanguage language, SpeechSynthesisVoiceName voiceName)
    {
        m_SpeechSynthesisLanguage = language.ToString().Replace('_', '-');
        m_SpeechSynthesisVoiceName = voiceName.ToString().Replace('_', '-');
    }
 
    /// <summary>
    /// 设置音源
    /// </summary>
    /// <param name="audioSource"></param>
    public void SetAudioSource(AudioSource audioSource)
    {
        m_AudioSource = audioSource;
    }
 
    /// <summary>
    /// 开始TTS
    /// </summary>
    /// <param name="spkMsg"></param>
    /// <param name="errorAction"></param>
    public void StartTTS(string spkMsg, Action<string> errorAction = null)
    {
        StopTTS();
        m_TTSCoroutine = StartCoroutine(SynthesizeAudioCoroutine(spkMsg, errorAction));
    }
 
    /// <summary>
    /// 开始TTS
    /// </summary>
    /// <param name="spkMsg"></param>
    /// <param name="audioSource"></param>
    /// <param name="errorAction"></param>
    public void StartTTS(string spkMsg, AudioSource audioSource, Action<string> errorAction = null)
    {
        SetAudioSource(audioSource);
        StartTTS(spkMsg, errorAction);
    }
 
    /// <summary>
    /// 暂停TTS
    /// </summary>
    public void StopTTS()
    {
        if (m_TTSCoroutine != null)
        {
            StopCoroutine(m_TTSCoroutine);
            m_TTSCoroutine = null;
        }
        if (m_AudioSource != null)
        {
            m_AudioSource.Stop();
            m_AudioSource.clip = null;
        }
    }
 
    public IEnumerator SynthesizeAudioCoroutine(string spkMsg, Action<string> errorAction)
    {
        yield return null;
        var config = SpeechConfig.FromSubscription(m_SubscriptionKey, m_Region);
        config.SpeechSynthesisLanguage = m_SpeechSynthesisLanguage;
        config.SpeechSynthesisVoiceName = m_SpeechSynthesisVoiceName;
        // Creates a speech synthesizer.
        // Make sure to dispose the synthesizer after use!
        using (var synthsizer = new SpeechSynthesizer(config, null))
        {
 
            // Starts speech synthesis, and returns after a single utterance is synthesized.
            var result = synthsizer.SpeakTextAsync(spkMsg).Result;
            //print("after   " + DateTime.Now);
 
            // Checks result.
            string newMessage = string.Empty;
            if (result.Reason == ResultReason.SynthesizingAudioCompleted)
            {
                // Since native playback is not yet supported on Unity yet (currently only supported on Windows/Linux Desktop),
                // use the Unity API to play audio here as a short term solution.
                // Native playback support will be added in the future release.
                var sampleCount = result.AudioData.Length / 2;
                var audioData = new float[sampleCount];
                for (var i = 0; i < sampleCount; ++i)
                {
                    audioData[i] = (short)(result.AudioData[i * 2 + 1] << 8 | result.AudioData[i * 2]) / 32768.0F;
 
                }
                // The default output audio format is 16K 16bit mono
                var audioClip = AudioClip.Create("SynthesizedAudio", sampleCount, 1, 16000, false);
                audioClip.SetData(audioData, 0);
                m_AudioSource.clip = audioClip;
                Debug.Log(" audioClip.length " + audioClip.length);
                m_AudioSource.Play();
 
 
            }
 
            else if (result.Reason == ResultReason.Canceled)
            {
                var cancellation = SpeechSynthesisCancellationDetails.FromResult(result);
                newMessage = $"CANCELED:\nReason=[{cancellation.Reason}]\nErrorDetails=[{cancellation.ErrorDetails}]\nDid you update the subscription info?";
                Debug.Log(" newMessage "+ newMessage);
                if (errorAction!=null) { errorAction.Invoke(newMessage); }
 
            }
 
 
        }
    }
}

3、AzureTTSStream

using UnityEngine;
using Microsoft.CognitiveServices.Speech;
using System.IO;
using System;
using System.Collections;
 
public class AzureTTSStream : MonoSingleton<AzureTTSStream>
{
    private AudioSource m_AudioSource;
 
    private string m_SubscriptionKey = "Your";
 
    private string m_Region = "Your";
 
    private string m_SpeechSynthesisLanguage = "zh-CN";
 
    private string m_SpeechSynthesisVoiceName = "zh-CN-XiaochenNeural";
 
    public const int m_SampleRate = 16000;
 
    public const int m_BufferSize = m_SampleRate * 60; //最大支持60s音频，但是也可以调大，流式的无所谓
 
    public const int m_UpdateSize = m_SampleRate / 10; //采样容量，越大越卡
 
    private Coroutine m_TTSCoroutine;
    private int m_DataIndex = 0;
 
    private AudioDataStream m_AudioDataStream;
 
    private void OnEnable()
    {
        StopTTS();
    }
 
    private void OnDisable()
    {
        StopTTS();
    }
 
    /// <summary>
    /// 你的授权
    /// </summary>
    /// <param name="subscriptionKey">子脚本的Key</param>
    /// <param name="region">地区</param>
    public void SetAzureAuthorization(string subscriptionKey, string region)
    {
        m_SubscriptionKey = subscriptionKey;
        m_Region = region;
    }
 
    /// <summary>
    /// 设置语音和声音
    /// </summary>
    /// <param name="language">语言</param>
    /// <param name="voiceName">声音</param>
    public void SetLanguageVoiceName(SpeechSynthesisLanguage language, SpeechSynthesisVoiceName voiceName)
    {
        m_SpeechSynthesisLanguage = language.ToString().Replace('_', '-');
        m_SpeechSynthesisVoiceName = voiceName.ToString().Replace('_', '-');
    }
 
    /// <summary>
    /// 设置音源
    /// </summary>
    /// <param name="audioSource"></param>
    public void SetAudioSource(AudioSource audioSource)
    {
        m_AudioSource = audioSource;
    }
 
    /// <summary>
    /// 开始TTS
    /// </summary>
    /// <param name="spkMsg"></param>
    /// <param name="errorAction"></param>
    public void StartTTS(string spkMsg, Action<string> errorAction = null)
    {
        StopTTS();
        m_TTSCoroutine = StartCoroutine(SynthesizeAudioCoroutine(spkMsg, errorAction));
    }
 
    /// <summary>
    /// 开始TTS
    /// </summary>
    /// <param name="spkMsg"></param>
    /// <param name="audioSource"></param>
    /// <param name="errorAction"></param>
    public void StartTTS(string spkMsg, AudioSource audioSource, Action<string> errorAction = null)
    {
        SetAudioSource(audioSource);
        StartTTS(spkMsg, errorAction);
    }
 
    /// <summary>
    /// 暂停TTS
    /// </summary>
    public void StopTTS()
    {
        // 释放流
        if (m_AudioDataStream != null)
        {
            m_AudioDataStream.Dispose();
            m_AudioDataStream = null;
        }
 
        if (m_TTSCoroutine != null)
        {
            StopCoroutine(m_TTSCoroutine);
            m_TTSCoroutine = null;
 
        }
        if (m_AudioSource != null)
        {
            m_AudioSource.Stop();
            m_AudioSource.clip = null;
            m_DataIndex = 0;
        }
    }
 
    /// <summary>
    /// 发起TTS
    /// </summary>
    /// <param name="speakMsg">TTS的文本</param>
    /// <param name="errorAction">错误事件（目前没有好的判断方法）</param>
    /// <returns></returns>
    private IEnumerator SynthesizeAudioCoroutine(string speakMsg, Action<string> errorAction)
    {
        var config = SpeechConfig.FromSubscription(m_SubscriptionKey, m_Region);
        config.SpeechSynthesisLanguage = m_SpeechSynthesisLanguage;
        config.SpeechSynthesisVoiceName = m_SpeechSynthesisVoiceName;
        var audioClip = AudioClip.Create("SynthesizedAudio", m_BufferSize, 1, m_SampleRate, false);
        m_AudioSource.clip = audioClip;
        using (var synthesizer = new SpeechSynthesizer(config, null))
        {
 
            var result = synthesizer.StartSpeakingTextAsync(speakMsg);
            yield return new WaitUntil(() => result.IsCompleted);
            m_AudioSource.Play();
            using (m_AudioDataStream = AudioDataStream.FromResult(result.Result))
            {
                MemoryStream memStream = new MemoryStream();
                byte[] buffer = new byte[m_UpdateSize * 2];
                uint bytesRead;
                do
                {
                    bytesRead = m_AudioDataStream.ReadData(buffer);
                    memStream.Write(buffer, 0, (int)bytesRead);
                    if (memStream.Length >= m_UpdateSize * 2)
                    {
                        var tempData = memStream.ToArray();
                        var audioData = new float[m_UpdateSize];
                        for (int i = 0; i < m_UpdateSize; ++i)
                        {
                            audioData[i] = (short)(tempData[i * 2 + 1] << 8 | tempData[i * 2]) / 32768.0F;
                        }
 
                        audioClip.SetData(audioData, m_DataIndex);
                        m_DataIndex = (m_DataIndex + m_UpdateSize) % m_BufferSize;
 
                        memStream = new MemoryStream();
                        yield return null;
                    }
                } while (bytesRead > 0);
 
            }
        }
 
        if (m_DataIndex == 0)
        {
            if (errorAction != null)
            {
                errorAction.Invoke(" AudioData error");
            }
        }
    }
}
 
/// <summary>
/// 添加更多的其他语言
/// 形式类似为 Zh_CN 对应 "zh-CN";
/// </summary>
public enum SpeechSynthesisLanguage
{
    Zh_CN,
}
 
/// <summary>
/// 添加更多的其他声音
/// 形式类似为 Zh_CN_XiaochenNeural 对应 "zh-CN-XiaochenNeural";
/// </summary>
public enum SpeechSynthesisVoiceName
{
    Zh_CN_XiaochenNeural,
}

4、MonoSingleton

using UnityEngine;
public class MonoSingleton<T> : MonoBehaviour where T : MonoBehaviour
{
    private static T instance;
    public static T Instance
    {
        get
        {
            if (instance == null)
            {
                // 查找存在的实例
                instance = (T)FindObjectOfType(typeof(T));
 
                // 如果不存在实例，则创建
                if (instance == null)
                {
                    // 需要创建一个游戏对象，再把这个单例组件挂载到游戏对象上
                    var singletonObject = new GameObject();
                    instance = singletonObject.AddComponent<T>();
                    singletonObject.name = typeof(T).ToString() + " (Singleton)";
 
                    // 让实例不在切换场景时销毁
                    DontDestroyOnLoad(singletonObject);
                }
            }
 
            return instance;
        }
    }
}

优化流式读取数据的卡顿

using FfalconXR;
using Microsoft.CognitiveServices.Speech;
using System;
using System.Collections;
using System.IO;
using System.Threading.Tasks;
using System.Threading;
using UnityEngine;
 
public class AzureTTSStream : MonoSingleton<AzureTTSStream>
{
    const string TAG = "【Unity AzureTTS】";
    bool m_IsShowLog=true;
 
    private AudioSource m_AudioSource;
 
    private string m_SubscriptionKey = "YOUR_KEY";
 
    private string m_Region = "eastasia";
 
    private string m_SpeechSynthesisLanguage = "zh-CN"; //"zh-cn-sichuan";
 
    private string m_SpeechSynthesisVoiceName = "zh-CN-XiaochenNeural"; // "zh-cn-sichuan-YunxiNeural"; //"zh-cn-shaanxi-XiaoniNeural";
 
    public const int m_SampleRate = 16000;
 
    public const int m_BufferSize = m_SampleRate * 60; //最大支持60s音频，但是也可以调大，流式的无所谓
 
    public const int m_UpdateSize = m_SampleRate / 10; //采样容量，越大越卡
 
    private Coroutine m_TTSCoroutine;
    private int m_DataIndex = 0;
    private SpeechConfig m_SpeechConfig;
    private AudioDataStream m_AudioDataStream;
    private Task m_Task;
    private CancellationTokenSource m_CancellationTokenSource;
    private SpeechSynthesizer m_SpeechSynthesizer;
    private bool m_IsSpeechSynthesizerInitialized;
    protected override void Awake()
    {
        base.Awake();
        m_SpeechConfig = SpeechConfig.FromSubscription(m_SubscriptionKey, m_Region);
        m_SpeechConfig.SpeechSynthesisLanguage = m_SpeechSynthesisLanguage;
        m_SpeechConfig.SpeechSynthesisVoiceName = m_SpeechSynthesisVoiceName;
    }
    private void Start()
    {
        InitializeSpeechSynthesizer();
    }
 
    private void OnEnable()
    {
        StopTTS();
    }
 
    private void OnDisable()
    {
        StopTTS();
    }
 
    protected override void OnDestroy()
    {
        StopTTS();
        DisposeSpeechSynthesizer();
    }
 
    public void SetAzureAuthorization(string subscriptionKey, string region) { 
        m_SubscriptionKey= subscriptionKey;
        m_Region= region;
    }
 
    public void SetLanguageVoiceName(SpeechSynthesisLanguage language, SpeechSynthesisVoiceName voiceName)
    {
        m_SpeechSynthesisLanguage = language.ToString().Replace('_','-');
        m_SpeechSynthesisVoiceName = voiceName.ToString().Replace('_', '-');
        Debug.Log(" m_SpeechSynthesisLanguage " + m_SpeechSynthesisLanguage);
        Debug.Log(" m_SpeechSynthesisVoiceName " + m_SpeechSynthesisVoiceName);
    }
 
    public void SetAudioSource(AudioSource audioSource)
    {
        m_AudioSource= audioSource;
    }
 
    public async void StartTTS(string spkMsg, Action startAction, Action<string> errorAction=null)
    {
        StopTTS();
        //m_TTSCoroutine = StartCoroutine(SynthesizeAudioCoroutine(spkMsg, startAction, errorAction));
        await SynthesizeAudioAsync(spkMsg, startAction, errorAction);
    }
 
    public void StartTTS(string spkMsg, AudioSource audioSource, Action startAction, Action<string> errorAction=null)
    {
        SetAudioSource(audioSource);
        StartTTS(spkMsg, startAction, errorAction);
    }
 
    public void StopTTS()
    {
        // 释放流
        if (m_AudioDataStream != null)
        {
            m_AudioDataStream.Dispose();
            m_AudioDataStream = null;
        }
 
        if (m_CancellationTokenSource != null)
        {
            m_CancellationTokenSource.Cancel();
        }
 
        if (m_TTSCoroutine != null)
        {
            StopCoroutine(m_TTSCoroutine);
            m_TTSCoroutine = null;
 
        }
        if (m_AudioSource!=null)
        {
            m_AudioSource.Stop();
            m_AudioSource.clip = null;
            m_DataIndex = 0;
        }
    }
 
    /// <summary>
    /// 发起TTS
    /// </summary>
    /// <param name="speakMsg">TTS的文本</param>
    /// <param name="errorAction">错误事件（目前没有好的判断方法）</param>
    /// <returns></returns>
    private IEnumerator SynthesizeAudioCoroutine(string speakMsg, Action startAction,Action<string> errorAction)
    {
        var audioClip = AudioClip.Create("SynthesizedAudio", m_BufferSize, 1, m_SampleRate, false);
        m_AudioSource.clip = audioClip;
        startAction?.Invoke();
        using (var synthesizer = new SpeechSynthesizer(m_SpeechConfig, null))
        {
            var result = synthesizer.StartSpeakingTextAsync(speakMsg);
            yield return new WaitUntil(() => result.IsCompleted);
            m_AudioSource.Play();
            using (m_AudioDataStream = AudioDataStream.FromResult(result.Result))
            {
                MemoryStream memStream = new MemoryStream();
                byte[] buffer = new byte[m_UpdateSize * 2];
                uint bytesRead;
                do
                {
                    bytesRead = m_AudioDataStream.ReadData(buffer);
                    memStream.Write(buffer, 0, (int)bytesRead);
                    if (memStream.Length >= m_UpdateSize * 2)
                    {
                        var tempData = memStream.ToArray();
                        var audioData = new float[m_UpdateSize];
                        for (int i = 0; i < m_UpdateSize; ++i)
                        {
                            audioData[i] = (short)(tempData[i * 2 + 1] << 8 | tempData[i * 2]) / 32768.0F;
                        }
 
                        audioClip.SetData(audioData, m_DataIndex);
                        m_DataIndex = (m_DataIndex + m_UpdateSize) % m_BufferSize;
 
                        memStream = new MemoryStream();
                        yield return null;
                    }
                } while (bytesRead > 0);
 
            }
        }
 
        if (m_DataIndex==0) {
            if (errorAction!=null)
            {
                errorAction.Invoke(" AudioData error");
            }
        }
    }
 
    /// <summary>
    /// async 方式
    /// </summary>
    /// <param name="speakMsg"></param>
    /// <param name="startAction"></param>
    /// <param name="errorAction"></param>
    private async Task SynthesizeAudioAsync(string speakMsg, Action startAction, Action<string> errorAction) {
        StopTTS();
        m_CancellationTokenSource = new CancellationTokenSource();
 
        if (!m_IsSpeechSynthesizerInitialized)
        {
            // 如果 SpeechSynthesizer 尚未初始化，则在需要时进行初始化
            InitializeSpeechSynthesizer();
        }
        try
        {
            using (var result = await m_SpeechSynthesizer.StartSpeakingTextAsync(speakMsg))
            {
                using (m_AudioDataStream = AudioDataStream.FromResult(result))
                {
                    var audioClip = AudioClip.Create("SynthesizedAudio", m_BufferSize, 1, m_SampleRate, false);
                    m_AudioSource.clip = audioClip;
 
                    startAction?.Invoke();
 
                    MemoryStream memStream = new MemoryStream();
                    byte[] buffer = new byte[m_UpdateSize * 2];
                    uint bytesRead = 0;
 
                    do
                    {
                        if (m_CancellationTokenSource.IsCancellationRequested)
                        {
                            break; // 取消操作，退出循环
                        }
 
                        bytesRead = await Task.Run(() => m_AudioDataStream.ReadData(buffer));
                        if (m_IsShowLog) Debug.Log(TAG + $"{bytesRead} bytes received.");
                        await memStream.WriteAsync(buffer, 0, (int)bytesRead);
                        if (memStream.Length >= m_UpdateSize * 2)
                        {
                            var tempData = memStream.ToArray();
                            var audioData = new float[m_UpdateSize];
                            for (int i = 0; i < m_UpdateSize; ++i)
                            {
                                audioData[i] = (short)(tempData[i * 2 + 1] << 8 | tempData[i * 2]) / 32768.0F;
                            }
 
                            audioClip.SetData(audioData, m_DataIndex);
                            m_DataIndex = (m_DataIndex + m_UpdateSize) % m_BufferSize;
 
                            if (m_AudioSource.isPlaying == false)
                            {
                                m_AudioSource.Play();
                            }
 
                            memStream = new MemoryStream();
                        }
 
                        await Task.Yield(); // 让出一帧的时间给主线程
 
 
                    } while (bytesRead > 0);
                }
            }
        }
        catch (Exception e)
        {
            if (m_IsShowLog) Debug.Log(TAG + e.Message);
        }
        
    }
 
    private void InitializeSpeechSynthesizer()
    {
        var config = SpeechConfig.FromSubscription(m_SubscriptionKey, m_Region);
        config.SpeechSynthesisLanguage = m_SpeechSynthesisLanguage;
        config.SpeechSynthesisVoiceName = m_SpeechSynthesisVoiceName;
 
        m_SpeechSynthesizer = new SpeechSynthesizer(config, null);
        m_IsSpeechSynthesizerInitialized = true;
    }
 
    private void DisposeSpeechSynthesizer()
    {
        // 停止语音合成并释放资源
        if (m_SpeechSynthesizer != null)
        {
            m_SpeechSynthesizer.StopSpeakingAsync();
            m_SpeechSynthesizer.Dispose();
            m_SpeechSynthesizer = null;
            m_IsSpeechSynthesizerInitialized = false;
        }
    }
}
 
public enum SpeechSynthesisLanguage {
    Zh_CN,
    zh_cn_sichuan
}
 
public enum SpeechSynthesisVoiceName {
    Zh_CN_XiaochenNeural,
}

附加：

声音设置相关

语言支持 - 语音服务 - Azure Cognitive Services | Azure Docs

中国部分声音选择设置：