Google Revolutionizes Translation with Gemini 3.5 Translate

Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Google Innovates with Gemini 3.5 Translate
With Gemini 3.5 Translate, Google aims to break down language barriers by translating conversations in real-time across more than 70 languages. This project echoes a long-held science fiction dream, seeking to enable almost instantaneous understanding between speakers of different languages. Gemini 3.5 Translate, also known as Gemini 3.5 Live Translate, stands out for its ability to provide real-time spoken translation.
Automatic translation is not a new concept, but Google claims to have reached a decisive milestone. Unlike older systems that waited for the end of a sentence to produce a translation, this new audio model processes and delivers the speech flow continuously, with only a few seconds of delay. This operation resembles the work of a human interpreter more than that of traditional software often compared to a tired GPS.
The tool will be integrated into Google Translate on Android and iOS, as well as in Google Meet. Developers will also have access to this technology via the Gemini Live API. Google is not just offering a practical feature for travel but is also laying the groundwork for international meetings, customer service, and mobility applications, where communication between speakers of different languages is essential.
A More Natural and Authentic Voice
One of the most remarkable aspects of Gemini 3.5 Translate is its ability to preserve the intonation, rhythm, and pitch of the original speaker's voice. Google promises a more natural translation experience, far removed from the usual mechanical voices. This innovation aims to prevent translations from sounding like a ticket dispenser at SNCF.
However, it remains to be seen how this technology will perform in real-world environments, often noisy and chaotic, such as a bustling restaurant or a hurried taxi. Google assures that it has trained its model to better handle ambient noise and automatically detect languages, thus sparing users from juggling language settings. No more need to manually change the setting whenever a speaker switches from Spanish to Japanese, which is both convenient and less disruptive during a conversation.
Additionally, Google is introducing SynthID, a digital watermark embedded in the audio signal to identify AI-generated content. This innovation raises ethical questions, as technology capable of mimicking a speaker's tone can be problematic. While translating is useful, making someone speak with a voice that is too convincing can become a slippery slope.
Google Meet: A Multilingual Space
Google Meet is one of the first services to benefit from this update. With Gemini 3.5 Translate, the platform can now handle over 2000 language combinations, allowing for direct exchanges between multiple languages without relying on English as an intermediary language. Previously, the system was much more limited and heavily depended on English as a pivot language.
For businesses, this means smoother international meetings, without makeshift subtitles, external interpreters, or long awkward pauses. However, challenges remain, such as accents, failed jokes, local expressions, and corporate phrases that even AI might want to overlook.
Google Translate also benefits from this update. With headphones, the translation can be streamed directly into users' ears. On Android, a new listening mode would even allow users to hold the phone to their ear like a traditional call to hear the translation more discreetly. This is a good idea to avoid sharing an entire conversation with nearby tables.
Partners like Grab are already testing the model to facilitate exchanges between drivers and passengers. This is likely where this type of tool can become most useful, in small everyday situations where a misunderstanding can lead to real complications.
One central question remains: real-time translation is not limited to words. It touches on context, nuances, and sometimes even the most basic diplomacy. Google is moving quickly, but to determine whether Gemini 3.5 Live Translate is truly a reliable translator or simply an excellent technological showcase, we will need to hear it in action in real situations.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.