音频与二进制转换

音频转二进制编码

import io
import base64

def generate(self, sentences: List[str], device='cpu'):
    assert isinstance(sentences, list) and isinstance(sentences[0], str), \
        'Input data should be List[str], but got {}'.format(type(sentences))

    paddle.set_device(device)
    audio_data = []
    for sentence in sentences:
        wav = self(sentence)
        with io.BytesIO() as buffer:
            sf.write(buffer, wav.numpy(), samplerate=self.samplerate, format='wav')
            audio_data.append(base64.b64encode(buffer.getvalue()).decode())

    logger.info('{} wave files have been generated'.format(len(sentences)))
    return audio_data

二进制转音频文件

import base64

def decode_audio_data(audio_data: str, output_path: str, samplerate: int):
    audio_binary = base64.b64decode(audio_data.encode())
    with open(output_path, 'wb') as f:
        f.write(audio_binary)
    return sf.read(output_path, dtype='float32')[0]