Ant Group open-sources Ming-3.0 voice models with five releases

Ant Group's Bailian team has open-sourced the Ming-3.0 voice series, with five models covering speech synthesis, audio understanding and dialogue, and speech recognition. Developers can pick models by task or combine them in their own workflows. The Ming-3.0-tts Voice Design feature lets users describe a target voice in text or clone it from a reference clip. A director mode lets creators explain how emotions shift, which words to stress, and how to handle pacing. Ming-3.0-audio-flash/tiny can transcribe, summarize and analyze long audio directly, keeping details like mood and speaker identity. Ming-3.0-asr-tiny improves recognition using data covering Mandarin, English, dialects and domain terms, and is already used in Alipay search and doctor assistant tools. All five models are now available on Hugging Face and ModelScope, with API services also offered.