6 second song like example but have this error

[wfq1971]

RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 568 but got size 440 for tensor number 1 in the list.

[16lkj]

The duration of the source audio and the target audio are not strictly aligned
You can consider adding the following code after the statement 'f0_ref = f0_ref * 2 ** (shift_semi / 12)' in the template_pitzer function

target_length = source_mel.shape[-1]
current_length = f0_ref.shape[-1]

if current_length != target_length:
    if current_length < target_length:
        f0_ref = np.pad(f0_ref, (0, target_length - current_length), 'constant', constant_values=0)
    else:
        pad_length = current_length - target_length
        min_val = source_mel.min()
        source_mel = np.pad(source_mel, ((0, 0), (0, pad_length)), 'constant', constant_values=min_val)