Imagine joining a Google Meet call with colleagues from all corners of the globe—one in Madrid, another in São Paulo, and you in Nairobi. You speak in English, your colleague responds in Spanish, and everyone hears the conversation in their own language, complete with the original speaker’s tone and emotion. This is the reality with Google Meet’s Gemini-powered real-time speech translation, set to transform cross-language communication.
While Google Meet has long provided real-time caption translation, the Gemini update takes things to a whole new level. Instead of merely reading what’s being said, you hear it—translated, in near real-time, and in a voice that closely resembles the original speaker. The result is a conversation that feels natural, dynamic, and unmistakably human.
How to access Gemini speech translation
Getting started is straightforward:
- On the web version of Google Meet, locate the “Speech translation with Gemini” panel in the top-right corner.
- Choose the language you’ll speak and the language you want to hear.
- Only one participant needs a Google AI Pro or Ultra subscription for everyone in the meeting to benefit.
A glowing Gemini overlay and a “Translating from [language]” indicator signal when the feature is active.
How Gemini makes it happen
At the core of this innovation is Gemini’s large language audio model, crafted by Google DeepMind. Here’s how it works:
- Input & Processing: As you speak, Gemini’s AI captures your voice and processes the audio instantly, translating your words into your conversation partner’s preferred language with minimal delay.
- Voice Preservation: Unlike traditional text-to-speech or robotic translators, Gemini dubs the translation using a synthetic voice that closely matches your own—tone, inflection, and emotional nuance included. It’s as if your multilingual counterpart is right there.
- Output: The listener hears a faint trace of your original speech, then the translated version takes over—smooth, expressive, and easy to follow.
This approach ensures your voice, tone, and expressions remain intact, even when translated, making conversations feel as natural as possible.
Current capabilities and supported languages
As of May 2025, Gemini-powered speech translation is in beta, supporting English and Spanish initially. Italian, German, and Portuguese are expected to be added soon, with more languages on the horizon.
Currently, the feature is available on the web version of Google Meet, with plans to expand to enterprise Workspace customers later in the year.
To use Gemini’s real-time speech translation, you’ll need:
- A Google Workspace subscription (Business Standard, Enterprise, or Gemini Education Premium), or
- A Google AI Pro or Ultra plan
Importantly, only one person in the call needs to be subscribed for everyone to enjoy the feature.
Performance, limitations, and real-world use
- Latency: There’s a slight delay as the AI processes and translates the speech, but Google is working to minimize this for a smooth conversation.
- Audio Quality: The original speaker’s voice is heard faintly before the translated voice takes over, maintaining context.
- Challenges: Gemini may occasionally struggle with regional dialects, slang, or strong accents. While it performs well in controlled settings, real-world accuracy can vary.
- Captions: While audio dubbing is the focus, captions can still be enabled for accessibility or review.
