Google Makes a New Translation Tool

Jun 9, 2026, 19:50
GlobalTechnology
A2

Google Makes a New Translation Tool

Google 推出的新翻譯工具

Introduction

Google has a new AI tool. It is called Gemini 3.5 Live Translate. It helps people speak different languages in real time.

Google 有一款新的 AI 工具，名為 Gemini 3.5 Live Translate，可幫助人們即時使用不同語言交流。

Main Body

In the past, you needed special Google headphones for this. Now, you can use any headphones. You can also use an Android phone. You hold the phone to your ear to hear the translation.

過去您需要使用 Google 專用的耳機才能達成。現在，您可以使用任何耳機，也可以使用 Android 手機，將手機貼近耳朵即可聽到翻譯。

This tool is very fast. It knows more than 70 languages. It works well in noisy places. The AI voice sounds like a real person. It does not sound like a robot.

這款工具速度非常快，支援超過 70 種語言，且在嘈雜環境中表現良好。AI 語音聽起來像真人，而不會像機器人。

Google is giving this tool to some people now. Soon, it will be in Google Meet. Later, everyone can use it in the Google Translate app. Google adds a special mark to the audio. This mark shows that an AI made the voice.

Google 目前正將此工具提供給部分使用者。不久後，它將整合至 Google Meet，隨後所有人都能在 Google 翻譯應用程式中使用。Google 會在音訊中加入特殊標記，以顯示該語音是由 AI 產生的。

Conclusion

Some people use the tool now. More people will use it soon.

目前已有部分人在使用，很快將會有更多人可以使用。

Vocabulary Learning

The 'Now' vs 'Soon' Shift

Look at how the text talks about time. This is the secret to moving from A1 to A2: connecting the present to the future.

1. Right Now (The Present)

  • Google is giving this tool...
  • Some people use the tool now.

2. In a Little While (The Future)

  • Soon, it will be in Google Meet.
  • More people will use it soon.

The Logic Path: Present Action →\rightarrow Future Result (Is giving) →\rightarrow (Will be)

Quick Tip: Using 'Soon' Put 'soon' at the end of your sentence to show something is happening very quickly in the future.

  • I am studying →\rightarrow I will speak English soon.
  • It is raining →\rightarrow The sun will come out soon.

Vocabulary Learning

translation (n.)
Changing words from one language into another language.
Example:I used a translation app to understand the Spanish menu.
real time (adj./adv.)
Happening now, without any delay.
Example:The app translates the speech in real time.
noisy (adj.)
A place with a lot of loud sound.
Example:It is hard to sleep in a noisy city.
robot (n.)
A machine that can do work automatically.
Example:The voice sounds natural and not like a robot.
B2

Google Launches Gemini 3.5 Live Translate Across Its Services

Google 在其服務中推出 Gemini 3.5 Live Translate 即時翻譯

Introduction

Google has introduced Gemini 3.5 Live Translate, a new AI model designed for speech-to-speech communication in many different languages in near real-time.

Google 推出了 Gemini 3.5 Live Translate，這是一款全新的 AI 模型，旨在實現多種不同語言之間近乎即時的語音對語音通訊。

Main Body

The development of Gemini 3.5 Live Translate marks a shift from relying on specific hardware to a more flexible software approach. In the past, real-time translation required specific Google devices, such as Pixel Buds. However, the new version allows users to use various earbud brands or a special 'listening mode' on Android devices, where the phone is held to the ear to hear translations.

Gemini 3.5 Live Translate 的開發標誌著從依賴特定硬體轉向更靈活的軟體方法。過去，即時翻譯需要特定的 Google 裝置，例如 Pixel Buds。然而，新版本允許使用者使用各種品牌的耳機，或在 Android 裝置上使用特殊的「聆聽模式」，將手機貼近耳朵即可聽到翻譯。

Technologically, the model is different because it uses continuous streaming instead of waiting for a speaker to finish a sentence. This system can automatically detect over 70 languages and reduces delays to only a few seconds. Furthermore, the system is designed to work in noisy environments by filtering out background sounds. To make the translation sound more natural, the AI tries to copy the speaker's original tone, speed, and pitch, which makes the voice sound less like a machine.

在技術上，該模型的不同之處在於它使用連續串流，而非等待說話者完成句子。此系統能自動偵測 70 多種語言，並將延遲降低至僅僅幾秒。此外，該系統設計用於在嘈雜環境中工作，可濾除背景雜音。為了讓翻譯聽起來更自然，AI 會嘗試模仿說話者的原始語調、速度和音高，使聲音聽起來不像機器。

Regarding its release, Google is launching the tool in stages. Developers can already access the model through the Gemini Live API or AI Studio. Additionally, business clients will see the tool integrated into Google Meet starting this month, followed by a general release in the Google Translate app for iOS and Android. To prevent the misuse of AI-generated audio, Google has added SynthID watermarks to all audio streams so that the AI speech can be easily identified.

關於發佈情況，Google 正分階段推出該工具。開發者已可透過 Gemini Live API 或 AI Studio 存取該模型。此外，企業客戶將從本月起在 Google Meet 中看到該工具的整合，隨後將在 iOS 和 Android 的 Google 翻譯 App 中全面發佈。為了防止 AI 生成音訊的濫用，Google 在所有音訊串流中加入了 SynthID 浮水印，以便能輕鬆識別 AI 語音。

Conclusion

Gemini 3.5 Live Translate is currently available to developers and some business users, while a wider release for the general public is expected soon.

Gemini 3.5 Live Translate 目前已向開發者及部分企業用戶開放，預計很快將向一般大眾全面發佈。

Vocabulary Learning

🚀 The "B2 Leap": Moving from Simple to Sophisticated Descriptions

At an A2 level, you describe things using simple verbs like is, has, or works. To reach B2, you need to describe processes and changes.

Look at this sentence from the text:

"The development of Gemini 3.5 Live Translate marks a shift from relying on specific hardware to a more flexible software approach."

💡 The Magic Phrase: "Marks a shift from X to Y"

Instead of saying "Things are different now," a B2 speaker says "This marks a shift from [Old Way] to [New Way]." It describes a transition perfectly.

Compare these two styles:

  • A2 (Basic): Google had special earbuds. Now, any earbuds work. It is a change.
  • B2 (Advanced): This new update marks a shift from using specific Google hardware to a more flexible software approach.

🛠️ Expanding Your "Connectors"

B2 speakers don't just use and or but. They use words that guide the reader. In the article, we see:

  1. "Furthermore" →\rightarrow Use this instead of "Also" when adding a strong new point.
  2. "Regarding..." →\rightarrow Use this to introduce a new topic (e.g., "Regarding the price, it is free.")
  3. "Instead of" →\rightarrow Use this to contrast two methods (e.g., "It streams audio instead of waiting for the end of the sentence.")

🎓 Pro-Tip: The "Less Like" Pattern

Notice how the author describes the voice:

"...which makes the voice sound less like a machine."

To sound more natural in English, stop using "It is not a machine" (which is too simple). Use "less like [Noun]" to describe a quality that is improving.

  • Example: "With practice, your English sounds less like a textbook and more like a native speaker!"

Vocabulary Learning

flexible (adj.)
Able to change or be adapted easily to different circumstances.
Example:The company adopted a flexible working schedule to help employees balance their home lives.
continuous (adj.)
Forming an unbroken whole; without interruption.
Example:The continuous noise from the construction site made it difficult to concentrate.
detect (v.)
To discover or identify the presence or existence of something.
Example:The new security system can detect motion even in complete darkness.
filter out (v. phr.)
To remove something unwanted from a group or a signal.
Example:These noise-canceling headphones filter out the sound of the airplane engine.
integrated (adj.)
Combined into a single unit or system.
Example:The new software provides an integrated platform for managing all project tasks.
misuse (n.)
The act of using something in the wrong way or for the wrong purpose.
Example:The government passed new laws to prevent the misuse of personal data.
identified (v.)
Recognized or established who or what someone or something is.
Example:The witness successfully identified the suspect in a police lineup.
C2

Google Announces the Integration of Gemini 3.5 Live Translate Across Its Ecosystem.

Google 宣布在整個生態系統中整合 Gemini 3.5 Live Translate 即時翻譯

Introduction

Google has introduced Gemini 3.5 Live Translate, a speech-to-speech AI model designed for near real-time multilingual communication.

Google 推出了 Gemini 3.5 Live Translate，這是一款專為近乎即時的多語言溝通而設計的語音對語音 AI 模型。

Main Body

The development of Gemini 3.5 Live Translate represents a transition from previous hardware-dependent iterations to a more versatile software implementation. Historically, real-time translation capabilities were contingent upon the utilization of specific Google hardware, such as Pixel Buds. The current iteration facilitates a broader deployment, permitting the use of diverse earbud brands or a dedicated 'listening mode' on Android devices, wherein the handset is held to the ear to receive translations.

Gemini 3.5 Live Translate 的開發代表了從先前依賴硬體的版本，轉型為更靈活的軟體實作。在過去，即時翻譯功能取決於是否使用特定的 Google 硬體，例如 Pixel Buds。目前的版本促進了更廣泛的部署，允許使用各種品牌的耳機，或在 Android 裝置上使用專用的「聆聽模式」，將手機貼近耳朵即可接收翻譯。

Technologically, the model diverges from traditional turn-based processing by employing continuous streaming translation. This mechanism allows for the automatic detection of over 70 languages and the mitigation of latency, with a delay of only several seconds. Furthermore, the system is engineered for environmental resilience, incorporating filters to neutralize background noise and overlapping vocalizations. To enhance the authenticity of the output, the model attempts to replicate the speaker's original intonation, pacing, and pitch, thereby reducing the prevalence of synthetic vocal characteristics.

在技術上，該模型捨棄了傳統的輪詢式處理，而採用持續串流翻譯。此機制可自動偵測 70 多種語言並降低延遲，延遲時間僅為數秒。此外，該系統旨在增強環境適應力，納入了可用於抵銷背景雜訊與重疊人聲的濾波器。為了提升輸出的真實感，該模型會嘗試複製講者的原始語調、節奏與音高，從而減少合成聲音的特徵。

Regarding institutional deployment, Google has initiated a phased rollout. Developers may currently access the model via the Gemini Live API or AI Studio. Enterprise clients will receive integration within Google Meet starting this month, followed by a general release within the Google Translate application for iOS and Android. To address potential concerns regarding synthetic media, Google has implemented SynthID watermarking within the waveform data of all audio streams to ensure the AI-generated nature of the speech remains identifiable.

關於機構部署，Google 已啟動分階段推出。開發者目前可透過 Gemini Live API 或 AI Studio 存取該模型。企業客戶將從本月起在 Google Meet 中獲得整合功能，隨後將在 iOS 與 Android 的 Google 翻譯應用程式中全面發布。為了處理對合成媒體的潛在疑慮，Google 在所有音訊串流的波形數據中實作了 SynthID 水印，以確保 AI 生成的語音仍可被識別。

Conclusion

Gemini 3.5 Live Translate is currently being deployed to developers and select enterprise users, with a wider consumer release pending.

Gemini 3.5 Live Translate 目前正部署給開發者與特定企業用戶，隨後將向更廣泛的消費者發布。

Vocabulary Learning

The Architecture of Nominalization and Precision

To transition from B2 to C2, a student must move away from verbal-centric descriptions (which often sound like storytelling) toward nominal-centric constructions (which sound like authoritative analysis). This text is a masterclass in Nominalization—the process of turning verbs or adjectives into nouns to create a denser, more objective academic tone.

⚡ The 'C2 Shift': From Action to Concept

Observe how the text avoids saying "Google changed how the software works" and instead opts for:

*"...represents a transition from previous hardware-dependent iterations to a more versatile software implementation."

Analysis:

  • Transition, Iterations, and Implementation are nouns derived from verbs (transition, iterate, implement).
  • By using nouns, the writer transforms a sequence of events into a set of established concepts. This removes the 'narrative' feel and replaces it with 'institutional' weight.

🔍 Linguistic Micro-Analysis: The Lexical Bridge

B2 Approach (Verbal/Direct)C2 Approach (Nominal/Abstract)Linguistic Effect
The system is resilient in loud places....engineered for environmental resilience.Shifts focus from the state of the system to the quality of the engineering.
It mitigates the delay....the mitigation of latency.Creates a formal object of study rather than a simple action.
The AI replicates the voice....attempts to replicate... reducing the prevalence of synthetic characteristics.Uses 'prevalence' (a noun of frequency) to quantify an abstract phenomenon.

🛠️ Stylistic Application: The 'Nominal Chain'

C2 writers often use "nominal chains"—sequences of nouns that specify complex ideas without needing multiple clauses.

Example from text: SynthID watermarking→waveform data→audio streams\text{SynthID watermarking} \rightarrow \text{waveform data} \rightarrow \text{audio streams}

Instead of saying "Google put watermarks into the data of the waves in the audio streams," the text chains the nouns to create a precise technical path. This density is the hallmark of native-level professional English.

Vocabulary Learning

contingent (adj.)
Dependent on certain circumstances; conditional.
Example:The success of the project is contingent upon the approval of the board of directors.
mitigation (n.)
The action of reducing the severity, seriousness, or painfulness of something.
Example:The new software update focuses on the mitigation of system latency during peak usage.
resilience (n.)
The capacity to recover quickly from difficulties or to withstand adverse conditions.
Example:The device was engineered for environmental resilience to ensure it functions in extreme weather.
neutralize (v.)
To render something ineffective or harmless by applying an opposite force or effect.
Example:The noise-canceling headphones are designed to neutralize ambient background sounds.
prevalence (n.)
The fact or condition of being common or widespread.
Example:The prevalence of synthetic vocal characteristics often makes AI voices sound unnatural.
phased (adj.)
Carried out in gradual stages rather than all at once.
Example:The company announced a phased rollout of the new feature to ensure system stability.
Practice All words in a crossword→