Compared with its predecessor Speech 02, MiniMax's new generation of speech generation model Speech 2.5 has made breakthroughs in three core metrics: multilingual expressiveness, timbre reproduction accuracy and language coverage. The model not only optimizes the generation of mainstream languages such as Mandarin Chinese and English, significantly improves the degree of speech similarity and natural rhythm, but also enhances the ability of cross-language timbre reproduction, which is able to accurately capture and preserve the details of the speaker's voice, including specific accents and speech intonation. In addition, Speech 2.5 adds support for niche languages such as Bulgarian and Danish, bringing the total number of supported languages to 40, which is invaluable for multilingual content deployment in globalized enterprises.
This answer comes from the articleMiniMax Releases Speech 2.5: Speech Synthesis Technology Breaks Through in Multilingualism and Tone ReproductionThe
English 
简体中文 
日本語 
Deutsch 
Português do Brasil