Hi,
I run latest HTS adaptation demo.
I adapt just 1 utt and generate samples with HMGENS.
I also adapt with 40 utts and create samples.
It seems there is no too much difference in quality.
Actually when i try slt_b500 (2.2639 seconds of speech) for adaptation, system can not adapt (output is like SI voice).
But when i use slt_b501.cmp (4.8762 second) system adapts well. But after that increasing the adaptation data does not help much.
I wonder how system can adapt that well with 1 utt.???