soulxsinger/utils/audio_utils.py

import torch
import torchaudio


def load_wav(wav_path: str, sample_rate: int):
    """Load wav file and resample to target sample rate.

    Args:
        wav_path (str): Path to wav file.
        sample_rate (int): Target sample rate.

    Returns:
        torch.Tensor: Waveform tensor with shape (1, T).
    """
    waveform, sr = torchaudio.load(wav_path)
        
    if sr != sample_rate:
        waveform = torchaudio.functional.resample(waveform, sr, sample_rate)

    if len(waveform.shape) > 1 and waveform.shape[0] > 1:
        waveform = torch.mean(waveform, dim=0, keepdim=True)
    
    return waveform