[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-microsoft-mai-transcribe-2-streaming":3,"topics-all":35,"news-related-96a256d8-cdca-4730-a566-a27985c5517c":54},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":21,"news_slug":28,"published_at":29,"created_at":30,"modified_at":31,"is_published":32,"publish_type":33,"image_url":14,"view_count":34},"96a256d8-cdca-4730-a566-a27985c5517c","微软语音三连发:流式转写百毫秒出稿,TTS 延迟 150 毫秒","微软发布三款语音模型:流式转写 MAI-Transcribe-2-Streaming 支持 60 语言,约 100 毫秒出初步转写,官方称登顶 AA 流式评测;MAI-Voice-2.1 覆盖 23 语言,Flash 版延迟 150 毫秒、每百万字符 15 美元,转写介绍价每小时 0.54 美元。","语音智能体有一个绕不开的物理约束:人耳对停顿的容忍度大概只有几百毫秒。听、理解、决策、说话,每一环都在这个时间窗里抢预算。10 月 1 日,微软一口气发布三款语音模型,把「听」和「说」两端的延迟同时往百毫秒级压——流式转写模型 MAI-Transcribe-2-Streaming、语音合成模型 MAI-Voice-2.1,以及为高并发场景准备的 MAI-Voice-2.1-Flash。\n\n## 先看转写:100 毫秒出「草稿」\n\nMAI-Transcribe-2-Streaming 是微软首个流式转写模型,支持 60 种语言,带自动连续语言检测。它不等说话人讲完再出全文:收到音频 100 毫秒出头就给出第一版「部分转写」(partials),随后随上下文推进不断修正,并很快锁定稳定文本。官方宣称,它在 Artificial Analysis 流式评测里,最终转写与部分转写的准确率双榜第一,并位于精度-延迟权衡的 Pareto 前沿;内部评测显示,实时听写场景下文字上屏速度约为最接近竞争对手的两倍。定价为介绍价每小时 0.54 美元,有效至年底。\n\n## 再看说话:一个声音走 23 种语言\n\nMAI-Voice-2.1 是微软目前最强的多语言 TTS:覆盖 23 种语言、26 个 locale,同一个「声音」可以先说英语、再切普通话和德语,每种语言都带原生口音,而不是把一种口音拖过所有语言。价格是每百万字符 22 美元。Flash 版面向高并发、延迟敏感负载:可生成 45 秒音频,端到端延迟仅 150 毫秒,推理比同档模型快 55%、便宜约 60%,每百万字符 15 美元。两个语音模型都支持用几秒参考音频跨全部支持语言克隆音色,并内置防滥用的同意防护。\n\n## 独立榜单交叉:第一之外还有第一\n\n只看官方口径容易被「第一」带着走。直接翻 Artificial Analysis 的公开榜单:非流式总榜上,MAI-Transcribe-2 词错率 2.0%,排第三——前两名是中国厂商的 StepAudio 3 ASR 与 Fun-Realtime-ASR-preview,都是 1.7%。微软的「双榜第一」限定在流式评测口径内。不过代际进步是真的:上一代 MAI-Transcribe-1.5 词错率 2.4%、速度因子 194;这一代 2.0%、372,同榜价格从每千分钟 6 美元降到 1.67 美元,降约七成,速度接近翻倍。速度榜第一仍是 Deepgram 的 Nova-3(578 倍速),微软排第二。\n\n## 所以呢:两端省下的毫秒,都是中间模型的推理时间\n\n微软在公告里把语音智能体描述成一个循环:听、理解、决策、说,每个组件要么帮你省时间,要么花你的时间。这次发布的真正卖点不是某个单项跑分,而是把循环两端的固定开销压到百毫秒量级之后,留给中间大模型去推理、调工具、自查答案的时间窗就宽了。三款模型已上线 Microsoft Foundry、MAI Playground 与 Vercel,语音模型另可通过 OpenRouter 调用,LiveKit 支持即将跟进。对做语音智能体的团队来说,转写每小时 0.54 美元的介绍价已经把「实时听」的成本打到接近白菜价——真正的瓶颈,回到了中间那个会思考的模型身上。\n\n参考:微软官方公告 https:\u002F\u002Fmicrosoft.ai\u002Fnews\u002Four-first-streaming-transcription-model\u002F ;Artificial Analysis 语音转写榜单 https:\u002F\u002Fartificialanalysis.ai\u002Fspeech-to-text","https:\u002F\u002Fmicrosoft.ai\u002Fnews\u002Four-first-streaming-transcription-model\u002F","2a8eff39-dd09-461e-a06f-35e87f40087e",[11,15,18],{"id":12,"name":13,"slug":13,"description":14,"color":14},"e676a5cf-1f24-472f-a765-86fa21a1bc3c","ai-model",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":19,"name":20,"slug":20,"description":14,"color":14},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",[22],{"id":23,"lang":24,"title":25,"summary":26,"content":27},"154c46aa-6cc7-4bf6-a7c1-dbf70f05cdeb","en","Microsoft ships voice trio: 100ms streaming ASR, 150ms TTS Flash","Microsoft launched MAI-Transcribe-2-Streaming (60 languages, ~100ms partials, $0.54\u002Fhour intro) with MAI-Voice-2.1 and Flash (150ms latency, $15 per 1M chars).","Voice agents live inside a hard physical constraint: human ears tolerate only a few hundred milliseconds of silence. Hearing, understanding, deciding, speaking — every component either buys time in that window or spends it. On October 1, Microsoft released three voice models that push both ends of the loop toward hundred-millisecond territory: the streaming transcription model MAI-Transcribe-2-Streaming, the TTS model MAI-Voice-2.1, and the high-volume variant MAI-Voice-2.1-Flash.\n\n## Transcription: first drafts in about 100ms\n\nMAI-Transcribe-2-Streaming is Microsoft's first streaming transcription model. It supports 60 languages with automatic, continuous language detection, and it does not wait for the speaker to finish: it emits its first partial hypotheses just over 100ms after receiving audio, then keeps revising them as context rolls in and commits a stable transcript promptly. Microsoft says it ranks no. 1 for accuracy on both final and partial transcripts in Artificial Analysis' streaming evaluation, sits on the accuracy-versus-latency Pareto frontier, and — per internal evaluations — puts words on screen 2x faster than its closest competitor in real-time dictation. Introductory pricing is $0.54 per hour of audio through the end of the year.\n\n## Speech: one voice across 23 languages\n\nMAI-Voice-2.1 is Microsoft's strongest multilingual TTS yet, covering 23 languages and 26 locales. A single voice can speak English, then Mandarin, then German, picking up a native accent in each rather than dragging one accent across languages. It is priced at $22 per 1M characters. The Flash variant targets high-volume, latency-sensitive workloads: it generates 45 seconds of audio with an end-to-end latency of just 150ms, runs 55% faster inference and is roughly 60% cheaper than comparable models, at $15 per 1M characters. Both voice models support voice cloning across all supported languages from a few seconds of reference audio, with built-in consent guardrails.\n\n## Cross-checking the leaderboard: whose \"no. 1\" is it?\n\nOfficial claims deserve a cross-check. On Artificial Analysis's public non-streaming leaderboard, MAI-Transcribe-2 posts a 2.0% word error rate — third place, behind two Chinese models, StepFun's StepAudio 3 ASR and Alibaba's Fun-Realtime-ASR-preview, both at 1.7%. Microsoft's \"no. 1 on both boards\" claim is scoped to the streaming benchmark. The generational jump is real, though: the previous MAI-Transcribe-1.5 scored 2.4% WER with a speed factor of 194, while the new model hits 2.0% at 372 — and its listed price on the same board dropped from $6.00 to $1.67 per 1,000 minutes, roughly 70% cheaper at nearly double the speed. The pure speed crown still belongs to Deepgram's Nova-3 at 578x real-time.\n\n## So what: milliseconds saved at the edges are reasoning time in the middle\n\nMicrosoft frames a voice agent as a loop — hear, understand, decide, speak — where every component either buys or spends time. The real story of this launch is not any single benchmark: once the fixed overhead at both ends drops to the hundred-millisecond scale, the window left for the model in the middle to reason, call tools, and check its answer gets wider. All three models are available through Microsoft Foundry, the MAI Playground and Vercel; the voice models are also on OpenRouter, with LiveKit support coming soon. For teams building voice agents, $0.54-per-hour transcription makes \"listening in real time\" nearly free — the real bottleneck has moved back to the model that does the thinking.\n\nReferences: Microsoft announcement https:\u002F\u002Fmicrosoft.ai\u002Fnews\u002Four-first-streaming-transcription-model\u002F ; Artificial Analysis speech-to-text leaderboard https:\u002F\u002Fartificialanalysis.ai\u002Fspeech-to-text","microsoft-mai-transcribe-2-streaming","2026-10-02T19:09:40Z","2026-10-02T19:09:46.438502Z","2026-10-02T19:09:46.438512Z",true,"agent",377,[36,45],{"slug":37,"tag_slug":37,"title_zh":38,"title_en":39,"intro_zh":40,"intro_en":41,"id":42,"is_active":32,"created_at":43,"modified_at":44},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":46,"tag_slug":46,"title_zh":47,"title_en":48,"intro_zh":49,"intro_en":50,"id":51,"is_active":32,"created_at":52,"modified_at":53},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":55},[56,61,66,71,76,81],{"id":57,"title":58,"news_slug":59,"published_at":60},"bbedb317-c221-43e3-b9a4-61798ed17b55","Voxtral 3:Mistral 语音智能体全栈,一轮对话不到一秒","voxtral-3-voice-agent-sdk","2026-10-21T23:11:24+00:00",{"id":62,"title":63,"news_slug":64,"published_at":65},"6062d551-9068-4a9f-ae8e-4e99269cd838","Muse Voice Transcribe 发布:流式转写、20+ 说话人分离、端点检测,Meta 全塞进一个模型","meta-muse-voice-transcribe-streaming-asr","2026-09-05T13:11:00+00:00",{"id":67,"title":68,"news_slug":69,"published_at":70},"c75518c3-da86-45a3-80fd-004e4650c06b","阿里把「世界模型」搬上百炼:HappyOyster 1.0 用一句话生成可实时交互的开放世界","alibaba-happyoyster-1-0","2026-07-19T22:00:00+00:00",{"id":72,"title":73,"news_slug":74,"published_at":75},"4c74f8b3-0a8d-4e00-9a7f-10ae44a3c004","Clairic抽取模型:不生成token,一遍前向出答案","cloudflare-clairic-extraction-model","2026-10-22T23:10:00+00:00",{"id":77,"title":78,"news_slug":79,"published_at":80},"35e9f682-7952-4f6a-a8e9-7f8bdd47e791","IBM 语音三件套更新:2B 旗舰 ASR 精度开源第二","granite-speech-5-collection-update","2026-10-22T21:08:48+00:00",{"id":82,"title":83,"news_slug":84,"published_at":85},"d83c4910-6dce-4171-8a44-9f68c9f841af","Qwen-Audio-3.1-ASR开源:转写自带说话人","qwen-audio-3-1-asr-speaker-diarization","2026-10-22T17:13:29+00:00"]