On the morning of October 30, MiniMax (Xiyu Technology) rolled out its newest speech model, namely MiniMax Speech 2.6. This model stands out with its end-to-end latency of under 250 milliseconds in audio generation. Moreover, it incorporates Fluent LoRA technology, which empowers users to clone voices and produce smooth, natural speech that aligns seamlessly with the intended text. (Here, “unveils” is a more vivid and commonly - used verb in English news for introducing new products compared to “launches”. “Rolled out” is also a more natural and idiomatic expression for introducing new things. “Stands out” better highlights the unique features of the model. “Clone voices” is a more accurate and commonly - used phrase in the context of voice replication. “Produce” is a more general and appropriate verb for creating speech, and “aligns seamlessly” gives a more vivid description of the matching degree between the generated speech and the target text.)