关于websocket 流式 TTS 服务 返回的音频base64

[syy-love]

请问返回的base64转成音频文件不成功~哪位大佬给看看
代码：
import base64
import io
import os
import soundfile
from paddlespeech.cli.log import logger
from paddlespeech.server.utils.audio_process import wav2pcm
from paddlespeech.server.utils.util import wav2base64

def postprocess(wav_base64: str, outfile: str) -> float:
audio_data_byte = base64.b64decode(wav_base64)
# from byte
samples, sample_rate = soundfile.read(
io.BytesIO(audio_data_byte), dtype='float32')

    # transform audio
    if outfile.endswith(".wav"):
        soundfile.write(outfile, samples, sample_rate)
    elif outfile.endswith(".pcm"):
        temp_wav = str(random.getrandbits(128)) + ".wav"
        soundfile.write(temp_wav, samples, sample_rate)
        wav2pcm(temp_wav, outfile, data_type=np.int16)
        os.remove(temp_wav)
    else:
        logger.error("The format for saving audio only supports wav or pcm")

报错：

[lym0302]

Soundfile first tries to load your system library. So put your compiled libsndfile somewhere on PATH, or in LD_LIBRARY_PATH, or in your current directory. Only when it can't find a system library will soundfile try its own.
来源：bastibe/python-soundfile#302