Speech-to-text technology has limitations that can be improved upon by considering various factors such as language bias, dialectal variations, and contextual nuances. To address these challenges, it's essential to approach languages with a flexible mindset and acknowledge the differences between languages, including their syntax, vocabulary, and evolution over time. By presenting models with diverse datasets and taking context into account, we can improve the accuracy of speech-to-text engines and deliver on our mission to understand every voice. Furthermore, being mindful of data bloat and its impact on training times and resources is crucial in optimizing our output while maximizing diversity.