ESL materials are often improved with the inclusion of high-quality recordings. Unfortunately, producing such recordings can be expensive, time-consuming, and complicated. Moreover, when recordings are included, they have typically been limited to Standard accents, even though non-Standard English is a natural part of ESL listening. Not only does this limit the students’ familiarity with non-Standard accents, but it also promotes accent discrimination.

AI speech models have improved dramatically and now have many expressive and naturalistic features. Moreover, these are available in a wide range of voices and languages with fine grained control over speed. Interestingly they can produce English speech in typical foreign accents. Finally, all of this is generally available to anyone producing language materials.

In this presentation, I introduce a set of these models, demonstrating what they can do and what we can do with them in ways that promote equity, diversity, inclusion and good language learning.