Many-to-Many conversion samples from unseen speakers

Description

We present the conversions generated by our method when converting unseen speakers i.e. not found in our method training data. For each conversion we display the source sample from the unseen speaker, the generated sample and finally a sample recorded by the target speaker.

Conversions

Source Speaker ID Target Speaker ID Source Sample Generated Sample Target Speaker Example Sample
260 8630
7127 6836
260 5514
260 5652
1580 3240
7176 374
2300 6415
2830 405
7729 2911
1320 7264
5683 7367
5142 40
4446 887
1089 7190
5683 2416
4992 587
4446 6531
3570 8312
7127 6081
8224 7226
5683 8088
6829 6367
6930 8108
4077 3168
7021 5104
8230 7800
237 3242
8555 8226
1580 5789
8463 696
2300 8095
2961 250
6829 254
1320 4859
3570 4830
3570 6000
8455 5867
237 412
7176 4195
1284 5163
2961 8609
2300 3664
5105 2893
5142 8747
2830 1116
4446 78
7176 8770
4992 8123
7021 2196