Brilliant! I was thinking about how well LLM's, STT and TTS models handle smaller languages, you know, where there's less than 10m or even 5m speakers worldwide. I know I can do a lot with languages with >100m speakers. My doubts about smaller languages comes from me talking Lithuanian (My native language ~4m speakers max worldwide) to ChatGPT - its not very good compared to larger languages. As smaller languages has a lot less data on the internet to train LLMs and other models on. Would you mind giving a short response here, whether it is a big challenge with smaller languages? (Probably would be useful for many people)