In actual production, it is recommended to do A/B listening tests on the same passage of text with different parameter combinations, and fix a set of parameters separately for different content types such as 'novel narration,' 'non-fiction narration,' and 'children's books,' to avoid drift in the listening experience across the whole book.