The Eleven v3 model showcases advancements in text-to-speech (TTS) technology, with capabilities that include singing and varying voice characteristics. Users have experimented with different prompts, discovering that while some songs prompted singing responses, others did not. The model allows for depth in instruction separation between spoken and specific outputs, leading to dynamic voice changing, although the production quality may vary. Comparisons were made with OpenAI's offerings, noting that while Eleven Labs has higher quality but is more expensive, OpenAI provides a more cost-effective solution with a less predictable output quality. While American English voices are praised, there's criticism regarding other language support and accent fidelity, indicating room for improvement and localization, particularly in non-English languages. The overall sentiment leans towards optimism for TTS capabilities where human-like interactions and persona are the goals, though issues with customer-service-like responses remain a concern.