如何在Python3.9中将Azure TTS音频输出至虚拟线缆供Discord使用？

步骤1：获取虚拟音频线缆的设备ID #

先运行以下代码枚举所有音频输出设备，找到虚拟线缆对应的设备ID：

import azure.cognitiveservices.speech as speechsdk

# 列出所有可用的音频输出设备
output_devices = speechsdk.audio.AudioDeviceInfo.enumerate_output_devices()
for idx, device in enumerate(output_devices):
    print(f"序号: {idx}, 设备名称: {device.name}, 设备ID: {device.id}")

在输出里找到名称包含“Virtual Cable”的设备，复制它的设备ID备用。

步骤2：修改TTS代码指定输出设备 #

在你的tts函数中，用获取到的设备ID创建AudioConfig，替代原来的占位符：

def tts(text):
    speech_config = speechsdk.SpeechConfig(subscription=os.getenv('SPEECH_KEY'), region=os.getenv('SPEECH_REGION'))
    speech_config.speech_synthesis_voice_name = "en-US-JaneNeural"
    # 替换成你复制的虚拟线缆设备ID
    audio_config = speechsdk.audio.AudioConfig(audio_output_device_id="你的虚拟线缆设备ID")
    speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=audio_config)

    speech_synthesizer.speak_text_async(text).get()

步骤3：优化原有代码的小问题 #

调整代码里的几个不合理之处，让运行更稳定：

• load_dotenv需要加括号调用，否则无法加载.env文件里的密钥
• 不要把函数定义放在while循环内部，移到循环外避免重复定义
• 增加识别结果判断，避免空文本传入TTS

修改后的完整代码：

import os
from dotenv import load_dotenv
import azure.cognitiveservices.speech as speechsdk

# 加载环境变量
load_dotenv()

def recognize_from_microphone():
    speech_config = speechsdk.SpeechConfig(subscription=os.getenv('SPEECH_KEY'), region=os.getenv('SPEECH_REGION'))
    speech_config.speech_recognition_language="en-US"

    audio_config = speechsdk.audio.AudioConfig(use_default_microphone=True)
    speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)

    print("Speak into your microphone.")
    speech_recognition_result = speech_recognizer.recognize_once_async().get()

    # 仅当识别成功时调用TTS
    if speech_recognition_result.reason == speechsdk.ResultReason.RecognizedSpeech:
        tts(speech_recognition_result.text)
    elif speech_recognition_result.reason == speechsdk.ResultReason.NoMatch:
        print(f"No speech could be recognized: {speech_recognition_result.no_match_details}")
    elif speech_recognition_result.reason == speechsdk.ResultReason.Canceled:
        cancellation_details = speech_recognition_result.cancellation_details
        print(f"Speech Recognition canceled: {cancellation_details.reason}")
        if cancellation_details.reason == speechsdk.CancellationReason.Error:
            print(f"Error details: {cancellation_details.error_details}")

def tts(text):
    speech_config = speechsdk.SpeechConfig(subscription=os.getenv('SPEECH_KEY'), region=os.getenv('SPEECH_REGION'))
    speech_config.speech_synthesis_voice_name = "en-US-JaneNeural"
    # 替换成你的虚拟线缆设备ID
    audio_config = speechsdk.audio.AudioConfig(audio_output_device_id="你的虚拟线缆设备ID")
    speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=audio_config)

    speech_synthesizer.speak_text_async(text).get()

# 循环执行识别
while True:
    recognize_from_microphone()

这样修改后，Azure TTS的音频会单独输出到虚拟线缆，不会影响其他软件的默认音频输出设备。

内容的提问来源于stack exchange，提问作者Arishia