88Z AI, 20 October — Ant Bailin has open-sourced the Ming-3.0 voice model series, comprising five models covering speech synthesis, audio understanding and dialogue, and speech recognition. These models can be flexibly combined for business workflows. The TTS model supports emotion and rhythm adjustment with paralinguistic tags, the audio model provides structured summaries of long recordings, and the ASR model supports hotword injection with an average error rate of 4.41%.
At a glance
- The Ming-3.0 series includes five models covering synthesis, understanding, and recognition
- TTS supports emotion adjustment, while ASR achieves a 4.41% average error rate

The update
Ant Bailin has officially open-sourced the Ming-3.0 voice model series, which includes five models covering three core tasks: speech synthesis, audio understanding and dialogue, and speech recognition. Developers can combine these models to fit their business workflows. Specifically, Ming-3.0-tts allows for fine-tuned emotion and rhythm adjustments via a director mode and supports inline paralinguistic tags. Ming-3.0-audio-flash can summarize and analyze long recordings, outputting structured results. Meanwhile, Ming-3.0-asr-tiny supports hotword and context injection, achieving an average error rate of 4.41% across four evaluation categories.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 百灵大模型Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.