Target sub-cost weights

[Morten H]

Is it possible to change the default target sub-cost weights for a multisyn voice? I. e. would it for example be possible to increase the weight of phonetic context or exclude stress from the taget cost calculation? (I assume that the features for the two parts of a diphone are equally weighted). Clark et al 2007 talks about this, but not about how to do it.

[Simon King]

Festival provides functions to change a variety of weights, including those within the join cost, but not those within the target cost (which are defined in code). See this post.

[Author]

Posts