开发基于Llama3.1、Whisper、OpenVoice v2的语音对话机器人时，在获取Llama3.1文本回复后调用OpenVoice v2转换语音，执行model.tts_to_file(text, speaker_id, src_path, speed=1.0)时出现错误：

OSError: [WinError 6] Handle is invalid

环境：Windows 11 + VSCode虚拟环境，已修复虚拟环境GPU不可用问题，文件路径确认无误。

相关代码：

def text_to_speech(text, reference_speaker, language='ja'):
    tone_color_converter, device, output_dir = init_openvoice()

    # Extract target speaker embedding from the reference speaker
    target_se, audio_name = se_extractor.get_se(reference_speaker, tone_color_converter, vad=True)

    # Use MeloTTS for base speaker text-to-speech generation
    model = TTS(language=language, device=device)
    speaker_ids = model.hps.data.spk2id
    
    base_speaker_key = list(speaker_ids.keys())[0]  # Use the first base speaker for simplicity
    speaker_id = speaker_ids[base_speaker_key]
    
    # Generate the base TTS audio
    src_path = f'{output_dir}/tmp.wav'
    model.tts_to_file(text, speaker_id, src_path, speed=1.0)
    
    # Run tone color conversion to match the reference speaker's voice
    save_path = f'{output_dir}/final_output.wav'
    source_se = torch.load(f'checkpoints_v2/base_speakers/ses/{base_speaker_key}.pth', map_location=device)
    
    tone_color_converter.convert(
        audio_src_path=src_path, 
        src_se=source_se, 
        tgt_se=target_se, 
        output_path=save_path,
        message="@MyShell"
    )

• 排查文件句柄占用：Windows下该错误多因文件未正确释放导致。先删除output_dir下的旧tmp.wav文件，避免残留进程占用；每次生成前可主动清理：

  import os
  if os.path.exists(src_path):
      os.remove(src_path)
  

• 显式指定设备：即使GPU已修复，MeloTTS可能存在设备识别偏差，手动指定设备测试：

  import torch
  device = 'cuda' if torch.cuda.is_available() else 'cpu'
  model = TTS(language=language, device=device)
  

• 替换tts_to_file为手动保存：绕开内置方法的文件句柄问题，直接生成音频数据再保存：

  import soundfile as sf
  # 替换原model.tts_to_file行
  audio_data = model.tts(text, speaker_id, speed=1.0)
  sf.write(src_path, audio_data, samplerate=model.hps.data.sampling_rate)
  

• 使用绝对路径：避免相对路径的解析问题，将路径转为绝对路径：

  src_path = os.path.abspath(os.path.join(output_dir, 'tmp.wav'))
  save_path = os.path.abspath(os.path.join(output_dir, 'final_output.wav'))
  

• 更新依赖：检查并更新torch、soundfile、MeloTTS等依赖包，修复版本兼容性问题。

内容的提问来源于stack exchange，提问作者정희준