deepvoice_client/local_plugins/azure_speech/ios/Classes/AzureTtsHelper.swift

import Foundation
import AVFoundation
import MicrosoftCognitiveServicesSpeech

/// Azure 语音合成辅助类
class AzureTtsHelper: NSObject {
    // 常量定义
    private let tag = "AzureTtsHelper"
    private let DEFAULT_LANGUAGE = "zh-CN"
    private let DEFAULT_VOICE = "zh-CN-XiaoxiaoNeural"
    
    // 核心组件
    private var synthesizer: SPXSpeechSynthesizer?
    private var speechConfig: SPXSpeechConfiguration?
    private var audioConfig: SPXAudioConfiguration?
    private var initialized = false
    private var speaking = false
    
    // 当前设置
    private var currentVoice = "zh-CN-XiaoxiaoNeural"
    private var currentRate = "0%"
    private var currentPitch = "0%"
    private var currentVolume = "100%"
    
    // 音频会话管理
    private let audioSession = AVAudioSession.sharedInstance()
    
    // 自定义音频输出流
    private var customAudioOutputStream: SPXPushAudioOutputStream?
    
    // 流式文本处理的缓冲区
    private var streamBuffer = ""
    private var lastSpeakTime: TimeInterval = 0
    
    // 应用上下文
    private let context: Any
    
    // 全局回调
    private var ttsCallback: TtsCallback?
    
    /// TTS回调接口
    protocol TtsCallback {
        /// 合成开始
        func onSynthesisStarted()
        
        /// 合成中
        func onSynthesizing()
        
        /// 合成完成
        func onSynthesisCompleted()
        
        /// 合成取消
        func onSynthesisCanceled()
    }
    
    /// 初始化
    /// - Parameter context: 应用上下文
    init(_ context: Any) {
        self.context = context
        super.init()
    }
    
    /// 初始化 TTS 引擎
    /// - Parameters:
    ///   - subscriptionKey: Azure 语音服务订阅密钥
    ///   - region: Azure 语音服务区域
    ///   - callback: 可选，全局TTS回调接口
    /// - Returns: 是否初始化成功
    func initialize(
        subscriptionKey: String,
        region: String,
        callback: TtsCallback? = nil
    ) -> Bool {
        print("\(tag): 初始化 Azure 语音服务")
        
        // 检查配置是否为空
        if subscriptionKey.isEmpty || region.isEmpty {
            print("\(tag): 错误: Azure 配置信息不完整")
            return false
        }
        
        // 释放之前的资源
        dispose()
        
        // 保存全局回调
        ttsCallback = callback
        
        do {
            // 创建语音配置
            speechConfig = try SPXSpeechConfiguration(subscription: subscriptionKey, region: region)
            
            // 设置语音合成输出格式为高质量音频
            speechConfig?.setSpeechSynthesisOutputFormat(.riff24Khz16BitMonoPcm)
            
            // 设置默认语音
            speechConfig?.setSpeechSynthesisVoiceName(currentVoice)
            
            // 创建音频配置
            if let customOutput = customAudioOutputStream {
                // 使用自定义音频输出流
                audioConfig = try SPXAudioConfiguration(streamOutput: customOutput)
            } else {
                // 使用默认扬声器
                audioConfig = SPXAudioConfiguration()
            }
            
            // 创建语音合成器
            synthesizer = try SPXSpeechSynthesizer(speechConfiguration: speechConfig!, audioConfiguration: audioConfig!)
            
            // 设置事件监听
            setupEventListeners()
            
            initialized = true
            print("\(tag): TTS 引擎初始化成功")
            return true
        } catch {
            print("\(tag): TTS 引擎初始化失败: \(error.localizedDescription)")
            return false
        }
    }
    
    /// 设置事件监听器
    private func setupEventListeners() {
        guard let synthesizer = synthesizer else { return }
        
        // 合成开始事件
        synthesizer.addSynthesisStarted { [weak self] _, _ in
            guard let self = self else { return }
            self.ttsCallback?.onSynthesisStarted()
        }
        
        // 合成中事件
        synthesizer.addSynthesizing { [weak self] _, _ in
            guard let self = self else { return }
            self.ttsCallback?.onSynthesizing()
        }
        
        // 合成完成事件
        synthesizer.addSynthesisCompleted { [weak self] _, _ in
            guard let self = self else { return }
            self.speaking = false
            self.ttsCallback?.onSynthesisCompleted()
        }
        
        // 合成取消事件
        synthesizer.addSynthesisCanceled { [weak self] _, _ in
            guard let self = self else { return }
            print("\(self.tag): 语音合成取消")
            self.speaking = false
            self.ttsCallback?.onSynthesisCanceled()
        }
    }
    
    /// 设置TTS回调
    /// - Parameter callback: TTS回调接口
    func setTtsCallback(_ callback: TtsCallback?) {
        ttsCallback = callback
    }
    
    /// 设置自定义音频输出流
    /// - Parameter outputStream: 自定义音频输出流，如果为nil则使用默认音频输出
    /// - Returns: 是否设置成功
    func setCustomAudioOutputStream(_ outputStream: SPXPushAudioOutputStream?) -> Bool {
        do {
            // 保存引用
            customAudioOutputStream = outputStream
            
            // 如果已初始化，需要重新创建合成器以应用新的音频输出流
            if initialized {
                recreateSynthesizer()
            }
            
            return true
        } catch {
            print("\(tag): 设置自定义音频输出流失败: \(error.localizedDescription)")
            return false
        }
    }
    
    /// 重新创建合成器
    private func recreateSynthesizer() {
        do {
            // 关闭现有合成器
            synthesizer = nil
            
            // 创建音频配置
            if let customOutput = customAudioOutputStream {
                // 使用自定义音频输出流
                audioConfig = try SPXAudioConfiguration(streamOutput: customOutput)
            } else {
                // 使用默认扬声器
                audioConfig = SPXAudioConfiguration()
            }
            
            // 使用新的音频配置创建合成器
            synthesizer = try SPXSpeechSynthesizer(speechConfiguration: speechConfig!, audioConfiguration: audioConfig!)
            
            // 重新设置事件监听
            setupEventListeners()
        } catch {
            print("\(tag): 重新创建合成器失败: \(error.localizedDescription)")
        }
    }
    
    /// 设置语音
    /// - Parameter voiceName: 语音名称，例如 "zh-CN-XiaoxiaoNeural"
    /// - Returns: 是否设置成功
    func setVoice(voiceName: String) -> Bool {
        if !initialized {
            print("\(tag): TTS 引擎尚未初始化")
            return false
        }
        
        if voiceName == currentVoice {
            print("\(tag): 已设置语音: \(voiceName)")
            return true
        }
        
        do {
            currentVoice = voiceName
            speechConfig?.setSpeechSynthesisVoiceName(voiceName)
            
            // 重新创建合成器
            recreateSynthesizer()
            
            print("\(tag): 已设置语音: \(voiceName)")
            return true
        } catch {
            print("\(tag): 设置语音失败: \(error.localizedDescription)")
            return false
        }
    }
    
    /// 设置语音合成参数
    /// - Parameters:
    ///   - rate: 语速，范围 -100 到 100，默认为 0
    ///   - pitch: 音调，范围 -100 到 100，默认为 0
    ///   - volume: 音量，范围 0 到 100，默认为 100
    /// - Returns: 是否设置成功
    func setSpeechParams(rate: Int = 0, pitch: Int = 0, volume: Int = 100) -> Bool {
        if !initialized {
            print("\(tag): TTS 引擎尚未初始化")
            return false
        }
        
        do {
            currentRate = formatPercentage(rate)
            currentPitch = formatPercentage(pitch)
            currentVolume = "\(min(max(volume, 0), 100))%"
            
            print("\(tag): 已设置语音参数: 语速=\(rate), 音调=\(pitch), 音量=\(volume)")
            return true
        } catch {
            print("\(tag): 设置语音参数失败: \(error.localizedDescription)")
            return false
        }
    }
    
    /// 格式化百分比值
    private func formatPercentage(_ value: Int) -> String {
        return value >= 0 ? "+\(value)%" : "\(value)%"
    }
    
    /// 生成 SSML 文本
    /// - Parameter text: 要转换的文本
    /// - Returns: SSML 格式的文本
    private func generateSsml(text: String) -> String {
        return """
        <speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="zh-CN">
            <voice name="\(currentVoice)">
                <prosody rate="\(currentRate)" pitch="\(currentPitch)" volume="\(currentVolume)">
                    \(text)
                </prosody>
            </voice>
        </speak>
        """
    }
    
    /// 合成文本为语音并播放
    /// - Parameter text: 要合成的文本
    /// - Returns: 是否成功开始合成
    func speakText(_ text: String) -> Bool {
        if !initialized {
            print("\(tag): TTS 引擎尚未初始化")
            return false
        }
        
        do {
            print("\(tag): 开始合成文本: \(text)")
            
            // 生成 SSML
            let ssml = generateSsml(text: text)
            
            // 标记为正在播放
            speaking = true
            
            // 异步合成语音
            try synthesizer?.speakSsmlAsync(ssml)
            
            return true
        } catch {
            print("\(tag): 语音合成异常: \(error.localizedDescription)")
            speaking = false
            return false
        }
    }
    
    /// 处理流式文本并在遇到标点符号时播放
    /// - Parameter text: 收到的文本流片段
    /// - Returns: 是否成功处理
    func speakStream(_ text: String) -> Bool {
        if !initialized || text.isEmpty {
            return false
        }
        
        do {
            // 添加新文本到缓冲区
            streamBuffer.append(text)
            
            // 增加500ms防抖逻辑
            let currentTime = Date().timeIntervalSince1970
            if currentTime - lastSpeakTime < 0.5 {
                return true
            }
            lastSpeakTime = currentTime
            
            let currentText = streamBuffer
            
            // 定义标点符号列表
            let punctuationMarks: [Character] = [".", "。", "!", "！", "?", "？", ";", "；", ",", "，", ":", "："]
            
            // 查找最后一个标点符号的位置
            var lastPunctuationIndex = -1
            for i in currentText.indices.reversed() {
                if punctuationMarks.contains(currentText[i]) {
                    lastPunctuationIndex = currentText.distance(from: currentText.startIndex, to: i)
                    break
                }
            }
            
            // 如果找到标点符号，则播放到该标点符号
            if lastPunctuationIndex >= 0 {
                // 提取要播放的文本（包含标点符号）
                let textToSpeak = String(currentText.prefix(lastPunctuationIndex + 1))
                
                // 剩余的文本保存在缓冲区中
                streamBuffer = String(currentText.dropFirst(lastPunctuationIndex + 1))
                
                // 播放提取的文本
                return speakText(textToSpeak)
            }
            
            // 如果没有找到标点符号，则等待更多文本
            return true
        } catch {
            print("\(tag): 流式语音合成失败: \(error.localizedDescription)")
            return false
        }
    }
    
    /// 播放剩余的流式文本
    /// - Returns: 是否成功播放剩余文本
    func flushStream() -> Bool {
        if !initialized {
            return false
        }
        
        do {
            // 获取缓冲区中剩余的文本
            let remainingText = streamBuffer
            
            // 清空缓冲区
            streamBuffer = ""
            
            // 如果缓冲区为空，直接返回成功
            if remainingText.isEmpty {
                return true
            }
            
            // 播放剩余文本
            return speakText(remainingText)
        } catch {
            print("\(tag): 刷新流式文本失败: \(error.localizedDescription)")
            return false
        }
    }
    
    /// 停止当前语音合成
    /// - Returns: 是否停止成功
    func stopSpeaking() -> Bool {
        if !initialized {
            print("\(tag): TTS 引擎尚未初始化")
            return false
        }
        
        do {
            // 清空流缓冲区
            streamBuffer = ""
            
            try synthesizer?.stopSpeakingAsync()
            speaking = false
            print("\(tag): 已停止语音合成")
            return true
        } catch {
            print("\(tag): 停止语音合成失败: \(error.localizedDescription)")
            return false
        }
    }
    
    /// 释放资源
    func dispose() {
        do {
            stopSpeaking()
            
            // 恢复音频会话
            try audioSession.setActive(false, options: .notifyOthersOnDeactivation)
            
            // 关闭自定义音频输出流
            customAudioOutputStream = nil
            
            // 清空流缓冲区
            streamBuffer = ""
            
            // 清理资源
            synthesizer = nil
            speechConfig = nil
            audioConfig = nil
            ttsCallback = nil
            
            initialized = false
            speaking = false
            print("\(tag): TTS 引擎已释放")
        } catch {
            print("\(tag): 释放 TTS 引擎失败: \(error.localizedDescription)")
            // 确保重置状态
            initialized = false
            speaking = false
        }
    }
    
    /// 检查当前是否正在播放语音
    /// - Returns: 是否正在播放语音
    func isSpeaking() -> Bool {
        return speaking
    }
}