| 37 | resample_rate = 16000 |
| 38 | |
| 39 | def do_resample(input_wav_path, output_wav_path): |
| 40 | """""" |
| 41 | print(f'input_wav_path: {input_wav_path}, output_wav_path: {output_wav_path}') |
| 42 | waveform, sample_rate = torchaudio.load(input_wav_path) |
| 43 | # 检查音频的维度 |
| 44 | num_channels = waveform.shape[0] |
| 45 | # 如果音频是多通道的，则进行通道平均 |
| 46 | if num_channels > 1: |
| 47 | waveform = torch.mean(waveform, dim=0, keepdim=True) |
| 48 | waveform = torchaudio.transforms.Resample( |
| 49 | orig_freq=sample_rate, new_freq=16000)(waveform) |
| 50 | utils_file.makedir_for_file(output_wav_path) |
| 51 | torchaudio.save(output_wav_path, waveform, 16000) |
| 52 | |
| 53 | |
| 54 | def do_decode(input_wav_path, input_prompt): |