AVSpeechSynthesizer.write方法标记回调未触发，如何解决？

核心原因：你使用的普通文本字符串AVSpeechUtterance默认不会生成任何语音标记，标记回调仅在合成器处理包含显式语音标记的话语时才会触发。

具体修正步骤 #

• 使用带语音标记的NSAttributedString创建话语，添加如AVSpeechSynthesisBoundaryMarkerAttribute这类标记属性；
• 或使用iOS 13+支持的SSML格式字符串，通过AVSpeechUtterance(ssmlString:)初始化话语。

示例代码（带边界标记） #

import AVFoundation

var synthesizer = AVSpeechSynthesizer()
synthesizer.mixToTelephonyUplink = false
synthesizer.usesApplicationAudioSession = false

// 创建带边界标记的富文本
let attributedString = NSMutableAttributedString(string: "Hello world")
// 在"Hello"结尾位置添加边界标记
attributedString.addAttribute(.AVSpeechSynthesisBoundaryMarkerAttribute, 
                              value: "hello_end", 
                              range: NSRange(location: 5, length: 0))

let utterance = AVSpeechUtterance(attributedString: attributedString)
utterance.voice = AVSpeechSynthesisVoice(language: "en-US")

synthesizer.write(utterance, toBufferCallback: { buffer in
    if let buffer = buffer as? AVAudioPCMBuffer {
        print("write \(buffer.frameCapacity > 1)")
    } else {
        print("could not create buffer or buffer empty")
    }
}, toMarkerCallback: { markers in
    print("markers received: \(markers)")
})

示例代码（SSML格式，iOS 13+） #

import AVFoundation

var synthesizer = AVSpeechSynthesizer()
synthesizer.mixToTelephonyUplink = false
synthesizer.usesApplicationAudioSession = false

// SSML中添加自定义标记
let ssmlString = """
<speak>
    Hello <mark name="hello_end"/> world
</speak>
"""

let utterance = try! AVSpeechUtterance(ssmlString: ssmlString)
utterance.voice = AVSpeechSynthesisVoice(language: "en-US")

synthesizer.write(utterance, toBufferCallback: { buffer in
    if let buffer = buffer as? AVAudioPCMBuffer {
        print("write \(buffer.frameCapacity > 1)")
    } else {
        print("could not create buffer or buffer empty")
    }
}, toMarkerCallback: { markers in
    print("markers received: \(markers)")
})

额外注意事项 #

• 确保使用的语音合成器语言与话语语言匹配，部分语音包可能不支持标记输出；
• 标记回调的触发时机对应合成到标记位置的时刻，若话语中无任何标记，回调永远不会执行。

内容的提问来源于stack exchange，提问作者Matt Bierner