Google announces Gemini 3.5 Transcribe for AI-powered speech-to-text

Google 发布 Gemini 3.5 Transcribe，助力 AI 驱动的语音转文字功能

While we wait (possibly in vain) for Gemini 3.5 Pro to launch, Google is releasing a different model in the 3.5 branch. The company has announced Gemini 3.5 Transcribe, an AI model designed to streamline voice input by editing out “ums” and corrections, outputting polished AI text. This model already powers the Gboard “Rambler” feature on the Pixel 11, but it’s about to appear throughout the Google ecosystem.

在我们等待（可能遥遥无期）Gemini 3.5 Pro 发布的同时，Google 在 3.5 系列中推出了另一款模型。该公司宣布了 Gemini 3.5 Transcribe，这是一款旨在简化语音输入体验的 AI 模型，它能自动剔除语音中的“嗯”、“啊”等口头禅及修正语，直接输出润色后的文本。该模型目前已为 Pixel 11 上的 Gboard “Rambler”功能提供支持，并即将推广至整个 Google 生态系统。

According to Google, Gemini 3.5 Transcribe is much faster and more accurate than its previous voice-to-text engine, known as Chirp 3. The new AI model should be about 70 percent faster from voice to final transcribed text, and the live-speech error rate has dropped to 5.5 percent. That’s only a little better than Chirp 3, which Google measures at 7.32 percent. Still, it’s a pain to fix typos when you’re using voice input, so any improvement here is beneficial.

据 Google 称，Gemini 3.5 Transcribe 比其上一代语音转文字引擎 Chirp 3 更快、更准确。从语音输入到最终文本生成，新模型的速度提升了约 70%，实时语音识别错误率降至 5.5%。虽然这仅比 Chirp 3 的 7.32% 错误率略有提升，但考虑到语音输入时修正错别字非常麻烦，任何改进都是有益的。

The new model isn’t just better at hearing the words you say; it’s also supposedly better at getting to the heart of what you meant to say. As you speak, Gemini 3.5 Transcribe can remove the awkward “ums” and “uhs” that clutter your stream of consciousness. It can also edit text on the fly if you need to correct yourself and refer to your provided custom vocabulary to handle “specialized jargon.” All this works in 85 languages and with up to three speakers in pre-recorded audio.

这款新模型不仅能更精准地识别你所说的词汇，据称还能更好地理解你的表达意图。当你说话时，Gemini 3.5 Transcribe 可以自动删除那些干扰思维连贯性的尴尬口头禅。如果你需要纠正自己的说法，它还能实时编辑文本，并参考你提供的自定义词汇表来处理“专业术语”。所有这些功能支持 85 种语言，并能处理预录音频中最多三人的对话。

The drawback, of course, is that you’re relying on the AI to accurately get the gist of your speech. For short blocks of text, the model seems good at cleaning up inconsistencies and verbal stumbles (based on my testing with Rambler), but the AI does technically change the wording of what you said, and that may not be appropriate for all situations.

当然，缺点在于你必须依赖 AI 来准确把握你的讲话要点。对于短文本，该模型在清理不连贯表达和口误方面表现良好（基于我对 Rambler 的测试），但从技术上讲，AI 会改变你所说内容的措辞，这在某些特定场合下可能并不适用。

Gemini 3.5 Transcribe is already live in a few places, including Rambler in Gboard. However, that’s still limited to Pixel 11 phones. Google says it will expand Rambler to “more Gemini Intelligence devices” later this year. If you’re using the Gemini app on macOS, your voice input will also benefit from Gemini 3.5 Transcribe starting today.

Gemini 3.5 Transcribe 目前已在部分场景上线，包括 Gboard 中的 Rambler 功能。不过，该功能目前仍仅限于 Pixel 11 手机。Google 表示，今年晚些时候会将 Rambler 扩展到“更多 Gemini 智能设备”上。如果你正在 macOS 上使用 Gemini 应用，从今天起，你的语音输入也将受益于 Gemini 3.5 Transcribe。

Developers will be seeing plenty of Gemini 3.5 Transcribe as well. Starting today, Antigravity will have the new model, with full access to screen context and chat history (with your permission). AI Studio’s build model also has Gemini 3.5 Transcribe now, allowing you to vibe code apps with AI-optimized voice-to-text. Developers can also access the model from the Gemini API.

开发者也将频繁接触到 Gemini 3.5 Transcribe。从今天起，Antigravity 将集成该新模型，并（在获得授权的情况下）完全访问屏幕上下文和聊天记录。AI Studio 的构建模型现在也已加入 Gemini 3.5 Transcribe，让你能够通过 AI 优化的语音转文字功能来编写代码应用。开发者还可以通过 Gemini API 访问该模型。

If you’re not using those specific apps or devices, you won’t see Gemini 3.5 Transcribe just yet, but it’s coming. Google says it plans to bring the AI transcription feature to the Chrome browser “soon.” When it’s available, you’ll be able to input text into any web field. Use it to write emails, leave angry comments, or even prompt AI chatbots by voice.

如果你目前没有使用上述特定的应用或设备，暂时还无法体验 Gemini 3.5 Transcribe，但它已经在路上了。Google 表示计划“很快”将该 AI 转录功能引入 Chrome 浏览器。届时，你将能够在任何网页输入框中使用它——无论是撰写邮件、发表评论，还是通过语音向 AI 聊天机器人发送指令。