From: Small-parallel exemplar-based voice conversion in noisy environments using affine non-negative matrix factorization
No. of training data
216
50
25
10
No. of frames
61,168
13,839
6782
2826