Search
As per the official announcement from Tongyi Large Models, Tongyi Bailing's speech large models, Fun-CosyVoice3 and Fun-ASR, have undergone substantial enhancements and are now available as open-source. Fun-CosyVoice3 has witnessed notable advancements in content coherence, speaker resemblance, and the natural flow of intonation. Remarkably, it can recreate the target voice using merely a 3-second audio clip, and it also supports dialect synthesis and tone customization. On the flip side, Fun-ASR has bolstered its contextual understanding and highly accurate speech transcription capabilities, leading to a more than 15% increase in recognition precision in specialized fields like home decoration and insurance. Additionally, it facilitates custom training for models exclusive to enterprises.
-
C114 Communication Network -
Communication Home
