tortoise-tts
tortoise-tts copied to clipboard
Keep getting this error.
What does this mean? Is it an issue with my source wav's?
audio.py:17: WavFileWarning: Chunk (non-data) not understood, skipping it.
I save them as 22,050k 32bit float from Audition.
Got the same error. I don't use python, but line 17 in audio.py is :
def load_wav_to_torch(full_path):
Which returns:
torch.FloatTensor(data.astype(np.float32)) / norm_fix, sampling_rate
So, if I'm reading this correctly, this error is originating somewhere in Torch.
Edit:
I was wrong, the error is coming from scipy.io.wavfile:
and to be more precise - the read function in scipy.io.wavfile which opesn a WAV file and returns "... the sample rate (in samples/sec) and data from an LPCM WAV file."
I'm having a hard time reading scipy.io.wavfile so I can't provide any further data.
I believe it means that your wav files have chunks in a format that the decoding program doesn't understand. I see it occasionally too and it doesn't seem to affect performance.