SPEECHSPLIT unofficial code demo

Reference paper:
Qian, K., Zhang, Y., Chang, S., Cox, D., & Hasegawa-Johnson, M. (2020). Unsupervised speech decomposition via triple information bottleneck. arXiv preprint arXiv:2004.11284.

Github code:
https://github.com/CODEJIN/SpeechSplit

Structure

VCTK 108 speaker version (106 trained, 2 unseen)

Result datasets

Speaker	Category	Source
VCTK P243	Consistent (Male)
VCTK P240	Consistent (Female)
VCTK P232	Trained(Male)
VCTK P277	Trained(Female)
VCTK P226	Unseen(Male)
VCTK P228	Unseen(Female)

Summary

Speaker		Source		Conversion
P243		P243 (Consistent)
		P232 (Trained, male)
		P277 (Trained, female)
		P226 (Unseen, male)
		P228 (Unseen, female)
P240		P240 (Consistent)
		P232 (Trained, male)
		P277 (Trained, female)
		P226 (Unseen, male)
		P228 (Unseen, female)

Combinations

Speaker	Rhyme	Content	Pitch