speaker ids does not match the samples generated.

#2
by mirfan899 - opened

I want to generate the sample in speaker_id0.mp3 voice, but when I pass the speaker id 0 to the model the generated voice is of male instead of female.

my code

with open("output/" + filename, 'wb') as f:
    with wave.Wave_write(f) as wav:
        model.synthesize(word, wav, speaker_id=0)

Here is the generated sample.

But I wanted to use this speaker.

https://huggingface.co./rhasspy/piper-voices/blob/v1.0.0/en/en_US/libritts/high/samples/speaker_0.mp3

This sounds like the sample rate in the WAV file is incorrect (maybe 16000 instead of 22050?)

Yes, changing the sample rate gives a different voice.

Sign up or log in to comment