2025: The Year Machine Interpreting Went Mainstream

With Zoom, Apple, and Google jumping in, real-time translation is becoming as ordinary as Wi-Fi.

Technologies rarely need to be perfect to change the world. They simply need to be everywhere. Smartphones, cloud storage, and video calls all followed this path: flawed at first, but once they became ubiquitous, their shortcomings mattered less than their accessibility. This year, speech-to-speech translation in general, and machine interpreting in particular, once a niche service for conferences and specialized platforms, is making the same leap.

The cascade of launches

All big names are adding momentum to this epochal change. At Zoomtopia 2025, Zoom just introduced real-time voice translation. Meeting participants can now speak and listen in their preferred language. This feature, available in a few months, adds to the external specialized translation apps already integrated in their ecosystem. For global companies, the consequences are immediate: the promise of smoother collaboration across continents, easier recruitment, and onboarding that no longer depends on shared language.

Google’s Pixel 10 lunched only a few weeks ago Voice Translate, enabling near-instantaneous translation during phone calls. The system preserves tone and runs entirely on-device, sidestepping privacy concerns. It works even if the other caller doesn’t own a Pixel. Ten languages are supported at launch, from French and German to Japanese and Spanish. A consumer feature for casual conversations. We tested it with a real call, you can see a video here in Niemdzi YouTube channel)

Not to be left behind, Apple embedded live translation in its AirPods, promising multilingual conversations on the go with the New York Times writing today that “Apple’s New AirPods Offer Impressive Language Translation“. Another consumer feature for casual conversations.

And beyond the giants, the market is filling fast. By some counts, 30 to 40 smaller and mid-sized apps, each serving local markets or specialized niches, have launched their own real-time translation features in 2025. They add to the big LSP specialized in this application for B2B customers, like KUDO and Interprefy. Major providers are also releasing ready-made APIs for developers: Microsoft has introduced a new speech-to-speech API, while newcomers like palabra.ai are quickly emerging as contenders. At the same time, many large language service providers (LSPs) are beginning to fold machine interpreting into their service portfolios, no longer treating it as competition but as an extension of their multilingual offerings.

Taken together, these moves signal a decisive shift. Machine interpreting is no longer a futuristic vision: it is becoming a standard feature, and for companies part of their infrastructure. Machine interpreting is becoming ubiquitous.

Imperfect, but transformative

The technology is far from flawless. Nuances are missed, accuracy wavers, and human interpreters still remain indispensable in high-stakes contexts, but quality is set to improve in the next few years (see my article on the super-human AI interpreter). Yet ubiquity changes expectations, and reality. Italian Philosopher Umberto Galimberti reminds us that when a phenomenon grows in quantity, it eventually transforms its quality. Heat added to water simply makes it warmer, until it reaches 100 °C. At that point, the same process produces a new state: steam. From a change of quantity (temperature), we reach a change of quality (the object itself). In other words, with billions of devices carrying speech translation, incremental launches are turning into a qualitative change in the way we communicate. The ecosystem of interpreting is changing.

Redefining the interpreting ecosystem

For the interpreting ecosystem, still sustained today by human professionals, the shift is disruptive. For interpreters, this disruption is not terminal, or a sign of immediate obsolescence. Machines, we speculate, will take on more casual and mid-level exchanges, while humans will remain indispensable in complex, trust-sensitive contexts. At least for the next few years. And this will remain the same also when highly capable machines will achieve performances that are on par with humans. In other words, the profession will not vanish, but it will be redefined. I believe that there are two things important recognizing at this moment: (i) highly capable machines and humans will share the stage of multilingual communication, each with their own strengths and limitations, and that (Ii) the interpreting ecosystem will be moving from a human-first to an AI-first paradigm (see also my article AI-First Interpreting Approach: What it is and Why it Matters). This has obviously great consequences for any stakeholders (change in the quality), also interpreters. Professional associations should recognize this, instead of continuing to avoid reality in the hope that that reality will not happen, and start to think how to reposition the service that they member offer.

The spread of machine interpreting also forces new and fascinating questions: How should systems be evaluated? Should these systems be regulated in some use cases? How? Who is accountable for errors in sensitive settings? What balance should be struck between convenience, security, and trust? The questions that stakeholders will need to navigate and answers are many.

Opportunities and pressures

For AI companies, ubiquity opens the market. Tech giants will dominate consumer-grade translation, but specialized enterprises can thrive in verticals: tailoring solutions for healthcare, legal, or government use; ensuring compliance and security; integrating seamlessly into enterprise workflows; and much more. For others, it will mean managing the transition and shaping a hybrid model in which human and AI interpreters coexist within their service portfolios.

Language service providers face sharp choices. They can build AI interpreting internally, partner with SaaS providers, or retreat to the shrinking but resilient niche of human-only services. In that case, survival depends on excellence, offering quality and trust no machine can match (for now). Each strategy brings its own risks and challenges.

What comes next

The year 2025 is likely to be remembered as the moment speech translation stopped being a novelty and became a default expectation. From boardrooms to phone calls, from classrooms to casual chats, language barriers are no longer assumed.

Like water turning to steam, ubiquity changes not just the scale of a phenomenon, but its nature. Machine interpreting is not replacing human intelligence; it is reshaping the ecosystem itself.