Official Pytorch Implementation for "DDDM-VC: Decoupled Denoising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust Voice Conversion" (AAAI 2024)
I've been reviewing the SynthesizerTrn Module within vc_dddm_mixup.py and it appears to be incomplete. Could you kindly provide the detailed implementation of the forward function? Your assistance would be greatly appreciated.
Additionally, I am encountering some difficulties in understanding the process described for randomly selecting the speaker style s_r using binary selection to achieve the mixed speech representation. This concept, referenced in the document on:
Page 2, Paragraph 3, Line 4
Section 4.3 on Page 6, Paragraph 2, Line 1
Page 7 under the Training section, Line 5
is crucial for my comprehension and application of the model. Could you please provide further clarification on this process or any example implementations?
Hello,
I've been reviewing the
SynthesizerTrnModule withinvc_dddm_mixup.pyand it appears to be incomplete. Could you kindly provide the detailed implementation of theforwardfunction? Your assistance would be greatly appreciated.Additionally, I am encountering some difficulties in understanding the process described for randomly selecting the speaker style s_r using binary selection to achieve the mixed speech representation. This concept, referenced in the document on:
is crucial for my comprehension and application of the model. Could you please provide further clarification on this process or any example implementations?
Thank you!