"The provided training data is very small for code verification purpose only"

[sbkim052]

Is this mean the pre-trained model given is kind of overfitted model trained on small dataset?

[auspicious3000]

no

[sbkim052]

Then is it possible to convert the voices of our own data with your pre-trained model? will it perform well?

[auspicious3000]

no, the pretrained model only works for speakers in the training set

[FurkanGozukara]

no, the pretrained model only works for speakers in the training set

How can I test my own speech (it is a wav file) data with your pre-trained model?

Lets say how can I train a model for the voice of trump and apply it to my speech?

I have read demo.ipynb and there is no input defined anywhere

I want to test my wav file with your pre-trained model

also i did everything and i am getting this error

PS C:\SpeechSplit> python demo.ipynb
Traceback (most recent call last):
  File "demo.ipynb", line 5, in <module>
    "execution_count": null,
NameError: name 'null' is not defined

@auspicious3000

[FurkanGozukara]

@sbkim052 can you check my problem? did you pass these steps?

https://github.com/auspicious3000/SpeechSplit/issues/28

[leijue222]

no, the pretrained model only works for speakers in the training set

@auspicious3000 Can I use it to train a model like Conan's bow-knot? Anyone's voice will be converted to the model's voice.

[leijue222]

Or this can only use the data in the training set, and can't convert the voice of a new speaker into the pre-trained model's voice.

The demo.ipynb can't customize the input voice and target voice, so I can't verify this.

[auspicious3000]

@leijue222 yes you can, but you need to re-train the model.

[leijue222]

@leijue222 yes you can, but you need to re-train the model.

That's wonderful. Actually, I want to use the baker's voice as Conan's bow-knot(Only one speaker), with your SpeechSplit to do voice conversion. Then, the SpeechSplit can convert my voice or your voice or anyone else voice to baker's voice.