Google 新 AI 转录功能可自动删除“呃”和“嗯”等语气词

Google 更新了 Gemini Audio，加入可自动识别专业术语以及超过 85 种语言的新转录能力。Gemini 3.5 Transcribe 是 Gemini 家族的新成员，紧随 3.5 Live Translate 发布；与此同时，Google 此前承诺将在 6 月推出的 Gemini 3.5 Pro 模型仍未发布。

Google 表示，3.5 Transcribe“相较于此前的转录模型 Chirp 3 实现了重大进步”，尤其体现在多语言表现和词语错误率方面。Google 称，该转录模型让用户能够“仅通过语音自然编辑”，并可自动格式化文本、移除“um”和“uh”等填充词。

用户可以向模型提供自定义词汇表，使 3.5 Transcribe 能自动适配特殊拼写要求和专业术语，从而避免手动编辑这些词语。它还可为预录音频中的最多三位说话人标注身份，并提供逐词时间戳。

除 3.5 Transcribe 外，Google 还表示，Gemini Audio 将于今日迎来 Gemini 3.5 Live 和 3.5 Live Experimental 更新，这些更新建立在为 Gemini 语音聊天模式提供支持的现有语音识别技术之上。本文发布后，Google 随后联系表示，这些额外模型暂未上线，也未提供新的发布时间。根据 Google 此前提供的信息，Gemini 3.5 Live 在处理中途打断、语言识别和实时视觉处理方面表现更佳；Gemini 3.5 Live Experimental 则更进一步，可在处理更复杂任务的推理过程中，实时逐步讲述其进展。

Gemini 3.5 Transcribe 从今天起开始向所有 macOS Gemini 应用英语用户推出，并通过 Android 上的 Rambler 听写功能面向部分国家和语言推出。开发者也可通过 AI Studio 和 Antigravity 中的 Gemini API 使用其公开预览版。Google 表示，对 Chrome 的支持即将推出。

更新，8 月 26 日： Google 在发布前向 The Verge 提供的信息中还提到了两个 Gemini 3.5 Live 和 3.5 Live Experimental 模型，但目前表示，今天仅发布 3.5 Transcribe。

Google has updated Gemini Audio with new transcription capabilities that automatically detect specialized jargon and more than 85 languages. Gemini 3.5 Transcribe is a new addition to the Gemini family that follows the launch of 3.5 Live Translate, and comes as we’re still waiting for Google to release the Gemini 3.5 Pro model that it promised to roll out in June.

Google says that 3.5 Transcribe “represents a major advancement from our previous transcription model, Chirp 3,” especially regarding multilingual performance and wording error rates. The transcription model allows users to “edit naturally with just your voice,” according to Google, and can automatically format text and remove filler words like “um” and “uh.”

Users can provide a customized vocabulary to the model, allowing 3.5 Transcribe to automatically adapt transcription to unique spelling requirements and specialized jargon to prevent those words from being edited manually. It can also attribute speech for up to three speakers in pre-recorded audio, alongside providing word-level timestamps.

Alongside 3.5 Transcribe, Google also said that 3.5 Live and 3.5 Live Experimental updates will be coming to Gemini Audio today that build on the existing speech recognition tech powering Gemini’s voice chat mode. After we published this story, Google then reached out to say these additional models arent being launched yet, and didn’t provide a new launch date. According to the information Google previously provided, Gemini 3.5 Live is better at handling mid-sentence interruptions, language recognition, and live visual processing, while Gemini 3.5 Live Experimental goes further by narrating its progress step by step in real time while it tackles reasoning on more complex tasks.

Gemini 3.5 Transcribe is rolling out starting today in English for all macOS Gemini app users, and the Rambler dictation feature on Android in select countries and languages. It’s also available for developers in public preview in the Gemini API via AI Studio and Antigravity. Google says that Chrome support is coming soon.

Update,August 26th: Google also mentioned two Gemini 3.5 Live and 3.5 Live Experimental in information provided to The Verge prior to publication, but now says that only 3.5 Transcribe is being announced today.