Hi All
I have built a HTS-2.2 voice for English
and synthesized some test samples with hts_engine_API with the built
models. What I keenly observed is, many a times the phoneme "AH" sounded
with higher duration and that is causing unnaturalness.
Is there any way to correct the (context dependent) duration
models of AH? (or) such an issue can be handled in any other way.
Thank you Sir,
- Anil