Fueling media coverage and creator tests and reviews on YouTube, tech giants Google, Amazon, Nvidia, Microsoft, and OpenAI had a busy month of May in 2025, sounding off at events about AI projects, products, and advancements that included a few notable language AI announcements.

One such announcement was the introduction of real-time, speech-to-speech translation within Google Meet, immediately available to users in English and Spanish after being launched and demoed during Google’s annual developer conference, Google I/O 2025, held on May 20-21, 2025.

The underlying technology is built on Google DeepMind’s AudioLM and Gemini AI. The audio model has been specifically trained for high-fidelity speech translation and aims for low-latency dubbing that simulates the presence of a live interpreter in the call (while a faint version of the original voice plays in the background for added clarity and context).

This functionality is rolling out in beta to users on the Google AI Pro plan, initially supporting English and Spanish, with plans for additional languages in the coming weeks.

The Translator API for Chrome extensions is now available alongside the Summarizer API and Language Detector API, which allows developers to integrate AI translation directly into web interfaces.

The Live API has also been significantly updated with Gemini 2.5 Flash Native Audio, offering developers granular control over voice, tone, speed, and style in agentic conversational AI applications in 24 languages. It can be tested in Google AI Studio by selecting “Stream.”

Google also announced SignGemma, a dedicated sign language model slated for release later in 2025 as part of the Gemma family of models. It was highlighted at the I/O event as the most capable model for translating sign language, especially ASL, into spoken language text to date.

Microsoft Build 2025, the company’s premier developer conference, took place from May 19-22, 2025 in Seattle. The event featured extensive content covering Visual Studio, VS Code, Azure,.NET, and all things AI for developers. 

New features announced within Azure AI Language are specifically designed to enhance and accelerate the development of AI agents. A new, dedicated Text Translation Agent Template, for example, allows developers to integrate text translation features directly into their AI agent workflows using Azure AI Translator.

Multilingual capabilities were presented as fundamental to the design and functionality of enterprise-grade AI agents. They enable the agents to communicate with customers in their preferred language and also manage translation requests. 

AI agents are expected to handle increasingly sensitive data across diverse linguistic contexts as well, so Microsoft addressed data privacy and compliance (e.g., GDPR, CCPA) by embedding personally identifiable information detection directly into multilingual translation pipelines.

Microsoft also introduced NLWeb, a new technology aimed at bringing conversational interfaces directly to the web. This allows users to interact with web content semantically, utilizing their choice of model and data. It is available as an open project on GitHub, so developers can choose to add multi-language support.

The Microsoft Build 2025 “Book of News” itself incorporates a translation button below the Table of Contents.

OpenAI Watch Party

There were no direct mentions of new translation or multilingual capabilities tied to specific announcements from OpenAI in May 2025. However, support for multiple languages is implied in the announcements of several international rollouts, starting with a product update addressing regional data governance requirements in Asia announced on May 7, 2025.

Other May 2025 language-relevant announcements at OpenAI were the launch of the “Operator” AI agent, on May 23, and the establishment of the “OpenAI Deutschland” (Germany) and “Stargate UAE” (United Arab Emirates) initiatives for the OpenAI for Countries Partnership.

NVIDIA Back on the World Stage

NVIDIA’s GTC Taipei 2025 was held from May 19-23, 2025, as part of the COMPUTEX 2025 event. The conference attracts tech leaders, developers, and researchers seeking to explore the latest breakthroughs in AI, robotics, and computing. 

GTC Taipei featured dedicated “Conversational AI Conference Sessions” designed to teach how to accelerate the entire pipeline of conversational AI, encompassing multilingual speech recognition, AI translation, and speech synthesis. 

Nvidia’s strategic focus is on providing the necessary AI infrastructure, computational power, and frameworks (like the NeMo Framework and the TRT LLM) for capabilities like real-time, multilingual conversational AI systems.

A specific session titled “Evaluations for Your Customer Assistant LLM Agent: No Time for Hallucinations” underscored the importance of evaluating LLMs for accuracy, reliability, and ethical use for local languages.

Amazon Q and Amazon QuickSight

Amazon placed a strong emphasis on the application of generative AI in business intelligence on a global scale with the Amazon Q and Amazon QuickSight Learning Series online sessions throughout May 2025. These sessions were offered with regional links (AMER, APAC, EMEA).

Amazon Q Developer had already announced expanded multilingual support in April for its integrated development environment and the Q Developer Command Line Interface. The expansion includes Mandarin Chinese, French, German, Italian, Japanese, Spanish, Korean, Hindi, and Portuguese. 

Signaling that the potential for decentralized innovation has never been greater, basically, all the technologies now available for developers from big tech include a multilingual AI component. There is also an emphasis across all companies on flexible AI frameworks to enable regional and language adaptations.

Next on the big tech global agenda is Apple’s Worldwide Developers Conference, which starts on June 9, 2025.