Imagine chatting with your abuela in Madrid or brainstorming with a teammate in São Paulo—without awkward pauses or the dreaded “Oops, I didn’t catch that.” That’s the promise Google just unveiled at its I/O 2025 conference: Gemini-powered, real-time speech translation baked right into Meet.

At the annual developer showcase, CEO Sundar Pichai hopped onto a Google Meet call and demonstrated the magic live on stage. An English speaker talked about weekend plans, and as they spoke, Meet seamlessly dubbed their words into Spanish—complete with their original inflections, laughter, and tone—so the Spanish-speaking colleague heard a natural, expressive translation instead of a robotic voiceover. Flip the roles, and Spanish became crisp, idiomatic English in real time. It felt less like a translation tool and more like an interpreter sitting beside you, smoothing every conversational bump.

Under the hood is a deep audio-language model—think AudioPaLM style—built by Google DeepMind and integrated into Gemini. Unlike caption-only translation, this system ingests your live audio, parses meaning, and synthesizes speech in the listener’s language with millisecond-level latency. You’ll still faintly hear the speaker’s original voice layered beneath the translation, so subtle paralinguistic cues—tone shifts, emphasis, even chuckles—aren’t lost.

By preserving the nuance of how you speak, Google aims to keep conversations feeling personal and spontaneous. That’s a big step beyond many existing solutions, which often deliver dry, flat translations that strip away emotion.

Right now, speech translation in Meet is rolling out in beta to consumers on Google’s AI Pro plan and to subscribers of the brand-new AI Ultra tier, which runs $249.99 per month. AI Pro subscribers get access to core Gemini features—like advanced search summarization and image generation—while AI Ultra lifts the ceiling, offering top-end models and priority compute for power users and enterprises.

According to Google, the Meet speech translation beta is hitting English–Spanish first, with support for Italian, German, and Portuguese slated to arrive “in the coming weeks.” Enterprise customers testing Workspace integration should see early previews later in 2025.

Microsoft Teams also dipped its toe into real-time translation earlier this year, previewing an AI translator that overlays translated captions—but it doesn’t yet preserve the speaker’s original voice. Teams’ approach leans on text pipelines: speech-to-text, machine translation, then text-to-speech in a generic AI voice. Google’s more ambitious pipeline keeps your vocal signature intact, aiming for a truly human-sounding exchange. Zoom has experimented with caption translation as well, but again, no voice dubbing—yet.

With every major calling platform racing to eliminate linguistic friction, we may soon take for granted that “language differences” no longer mean “communication barriers.”

Real-world use cases

• Family calls across generations: Grandchildren can speak English, grandparents reply in Spanish—without switching apps or stressing over subtitles.
• Global teams: Companies with distributed teams can brainstorm naturally, even if no one shares a common mother tongue.
• Customer support: Support reps could help callers in dozens of languages while maintaining a personal tone.

Early testers praise the low latency—translations typically appear within a fraction of a second, making multi-person conversations flow smoothly.

Google’s roadmap hints at broader language support soon, and deeper Workspace integration—think translated meeting transcripts, AI-generated meeting minutes in your language of choice, even on-the-fly glossary creation for domain-specific terms. Down the line, we might see augmented-reality glasses overlaying real-time subtitles in your field of view, or mobile Meet calls auto-translate at the edge with zero cloud hops.

Real-time speech translation isn’t just a flashy demo—it’s a tangible leap toward a world where language is no longer a hurdle to human connection. By preserving the subtle rhythms of our voices, Google is betting that technology should amplify our humanity, not erase it. For anyone who’s ever paused mid-call, stumbling through a foreign phrase, this could be the start of effortlessly natural multilingual conversation—one Gemini translation at a time.

Stay tuned as this beta rolls out, and get ready to say “¡Hola!” or “Ciao!” without missing a beat.