Ant Group open-sources 5 Ming-3.0 speech models
The three links are composable and self-deployable, so a team can build a full listen-understand-speak voice agent without paying for proprietary APIs. Deployers in multilingual markets gain a working path for dialect-heavy ASR and instruction-driven TTS in one stack. Compared with closed services like OpenAI's audio APIs, this stack keeps weights local and lets you fine-tune each stage separately.
Ant Group (蚂蚁集团) released 5 open-source Ming-3.0 models covering TTS, native audio understanding and ASR. Ming-3.0-tts supports director-style voice generation, controlling tone, speed, pitch, emotion and paralanguage like laughter via natural-language instructions, in Tiny 7.9B/A1.3B and 0.5B sizes. Ming-3.0-audio understands and reasons over raw audio without transcription, answering in text or speech, in Flash 126B-A6.5B and Tiny 9.3B-A2.7B versions. Ming-3.0-asr-tiny improves recognition of dialects and professional terms with Context ASR using hotwords and conversation history.
蚂蚁一口气又开源出来了5个模型，涵盖了听懂、回答、说话三个环节，TTS、原生音频理解、ASR
可组合、可自行部署，用这个链路可以搭一个从“听—理解—合成”的语音Agent出来
TTS方面：Ming-3.0-tts，最有价值的是“导演式语音生成”，声音可以通过自然语言来设计
模型可以按照指令控制音色、语速、音高、情绪、重音以及叹气、笑声、停顿等副语言表现
有两款模型：Tiny 7.9B/A1.3B、0.5B
Audio方面：Ming-3.0-audio，不是单纯的ASR也不是TTS，应该说是有听觉能力的大模型，原生音频理解与推理，无需中间转写
能直接听懂声音、理解和推理音频，以文字或语音回答，也可以结合外部工具构建语音Agent
比如说总结长录音、回答口头问题、理解音乐和环境声音等
也是两款模型：Flash 126B-A6.5B/Tiny 9.3B-A2.7B
ASR方向：Ming-3.0-asr-tiny，这次重点强化了方言、专业术语以及上下文识别
核心是它的Context ASR能力，结合专业术语、候选热词和历史对话，可以提高复杂场景下的识别准确率
#ASR #TTS #蚂蚁百灵大模型
