[FastPitch1.1/pytorch] pitch transform lead to noisy speech

[JohnHerry]

Related to FastPitch1.1/pytorch

Is your feature request related to a problem? Please describe. A question about FastPitch1.1/pytorch inferene, there are some command arguments, like --pitch-transform-flatten; --pitch-transform-invert; --pitch-transform-amplify. I had listened the samples in FastPitch demo page, they are good. But in my instance, this inference arguments would create noisy speech but not the desired. My model is trained on two speakers, the pitch-std, pitch-mean are compuated from all samples together. I found that the predicted pitches are mostly positive numbers, So I don't know how did the --pitch-transform-invert generate good speech in the demo page? Is there any mistake that I may had done?

Describe the solution you'd like A clear and concise description of what you want to happen.

Describe alternatives you've considered A clear and concise description of any alternative solutions or features you've considered.

Additional context Add any other context or screenshots about the feature request here.