facebook

/

mms-tts-hin

like 20

Follow

AI at Meta 13.1k

So I was using this model to perform TTS-hindi but while saving the audio files I faced some errors. So I have resolved the error. If anyone is also facing any error while saving the audio file then you can try this function. It worked for me.

def save_audio_file(output, file):
audio_array = output.numpy().squeeze()
audio_array /=1.414
audio_array *= 32767
audio_array = audio_array.astype(np.int16)

sampling_rate = model.config.sampling_rate

scipy.io.wavfile.write(f"/root/{file}", rate=sampling_rate, data=audio_array)