Google brings its AI to more than 300 languages

Google says its products and technologies now work in more than 300 languages and reach more than 7 billion people, close to 86% of the world's population. The company presented this progress as part of its plan to help artificial intelligence better understand how people speak, even when they mix languages, use local dialects or communicate through sign language.

The figure does not mean that all those languages receive the same level of support. Google acknowledges that thousands of languages remain underrepresented online and in AI systems. That is why its next goal is to work with the 1,000 most widely spoken languages in the world, including many with limited digital data.

From translating words to understanding conversations

Traditional voice systems usually follow three steps: they convert audio into text, process that text and then generate a spoken response. This method works, but it can lose important information such as tone, pauses, emotion or context.

Google's new models attempt to process audio directly. This allows them to interpret less orderly conversations, including laughter, interruptions, hesitation or switches between languages in the middle of a sentence, as happens with Spanglish or Hinglish.

The tools announced or highlighted by Google include:

• Gemini 3.5 Live Translate, which provides real-time spoken translation in 70 languages and more than 2,000 language combinations.
• Gemini 3.5 Transcribe, designed to convert audio into clean, structured text even with background noise or specialized vocabulary.
• Rambler, a Gboard feature that removes filler words, corrects grammar and punctuation, and lets you edit text through voice commands.

In practice, this could help you hold a conversation with someone who speaks another language, dictate a message without worrying about punctuation or transcribe a meeting with technical terms.

Data collected with local communities

To improve performance in languages with limited digital resources, Google is funding projects that collect recordings and other data directly with local communities. The company highlights three initiatives:

• WAXAL gathers voice data from 27 Sub-Saharan African languages spoken by more than 100 million people in more than 26 countries.
• Project Vaani, developed with the Indian Institute of Science and Bhashini, has collected more than 30,000 hours of voice data in 109 languages in India, with the participation of more than 155,000 speakers.
• Amplify Initiative has worked with more than 1,600 local experts and 20 universities across four continents to gather 15,000 multimodal data samples.

Google also introduced Language Explorer, a tool that visualizes LinguaMeta, an open repository containing information about more than 7,000 spoken, written and signed languages.

AI without a connection and on basic phones

Internet access remains a barrier for more than 3 billion people, according to data cited by Google. For those cases, the company developed TranslateGemma, a family of lightweight translation models trained in 55 languages that can run directly on a device without a constant connection to the cloud.

Hardware is another problem. Hundreds of millions of people use basic phones that cannot run demanding models. That is why Google supports Viamo, whose AVA voice assistant lets people ask Gemini questions from a conventional phone. In a test in Rwanda, the service has already answered more than 2 million questions.

More accessibility and more cultural context

The plan also includes tools for people whose forms of communication are often poorly understood by voice systems. Sign Language-to-Text technology, or SL2T, has been trained on more than 50 sign languages and can convert signs into text in features such as Gboard and Live Transcribe, starting with American Sign Language.

Google estimates that around 70 million people use sign languages to communicate. The company is also working with local specialists to improve the pronunciation of street and place names. In New Zealand, for example, it worked with Māori language experts so that Google Maps pronounces place names more accurately.

For you, the most visible change may appear in everyday features: more natural translations, more accurate voice dictation, apps that work without an internet connection and better results in languages that have received little attention until now. But the challenge will not be solved simply by expanding the number of supported languages. What matters will be whether AI understands their nuances, respects how each community communicates and maintains that level of quality outside controlled demonstrations.